Anton Antoniv - S1E10

What Does Your Resume Look Like to a Computer

Anton is an applied mathematician (PhD) with 28+ years of experience in algorithm development, scientific computing, mathematical modeling, natural language processing, combinatorial optimization, research and development programming, machine learning, and data mining. In the last 11 years, he focused on developing machine learning algorithms and workflows for different industries (entertainment, recruitment, healthcare, manufacturing.) He is a former kernel developer of Mathematica.

Some would say AI is the snake oil of the 21st century. It's being marketed to you in ways like never before. If you're going to go buy something because of its AI capabilities, you owe it to yourself to really know what you're being sold. In this episode, Dave talks with Anton about what AI’s real capabilities are and if it is or isn’t really artificial intelligence.

Anton joins Dave to talk about:

  • Artificial Intelligence
  • Strong AI vs Weak AI
  • How AI is applicable in recruiting
  • How to outsmart the ATS (spoiler alert: you can’t. The best thing to do is be your authentic self)
  • ...and much more!
RecruiterCast on Apple Podcasts
RecruiterCast on Google Podcasts
RecruiterCast on Spotify
RecruiterCast on Stitcher Podcasts

Being an immigrant, Anton compares and contrasts the U.S. labor market to European labor markets and shares some interesting insight. He also discusses how your resume may inadvertently cause discrimination and provides some tips to recruiters and employers on how they can avoid this. He discusses the importance of knowing your audience for your resume – who will be reviewing your resume? What generation are they in? These two questions can help you present a resume that will help you land the job.

"Artificial intelligence, I just don't believe in this. I think it's a great marketing term."
- Anton Antonov
"The best way for the computer, the applicant tracking system and the machine learning to find you is for you to authentically represent who you are and what you've done and what it is that you would like to do on your resume."
- Dave Webb
"Knowing your audience for your resume, it's very important, and ideally, you should cater to that."
- Anton Antonov

Links Associated with this Episode:

Show Notes:

Dave Webb:                           00:02              "Hi RecruiterCast listeners. It's Dave, your host and I just need 60 seconds of your time before we get into the show today. I'd just like to thank you so much for listening. We have great download numbers. We're getting great feedback from the community and the recruiting industry and we just like to say please keep sharing with your recruiter friends or any friends for that matter. We are recording season two already. We have a great guest lineup for you coming up in season two. So, there're a couple spots left, if you'd like to be a guest, go to and fill out the form. Also, we're going to do a QA round table here before season one ends and we'd love to feature some of your questions. You can email those questions to or call us at (904) 525-8134 just leave a message and we will get your voice on the air. Thank you for listening and here's your weekend quotes."

Music:                               00: 55

Dave Webb:                           00:56                  "I'm Dave Webb and you're listening to RecruiterCast."

Speaker 3:                           01:03                  "Data science, what is dealing with, it's a amalgam of computer science tasks, data engineering, collect software engineering, data engineering tasks and statistics. It's a strong overlap. It's an overloaded term , different people understand different things. If you say data scientist, it's a, from my perspective, data scientists, it's like a, it's not my full moldation, but they're the best a software engineer amongst statistician. They're the better programmers in the mathematics scientific communities and they're the better statistician or the better mathematicians when it comes to computer scientists."

Dave Webb:                           01:44                   "That is Anton Antonov, a senior data scientist. Now before we get too far into this, please know this is going to be a very technical episode. Anton is a very smart guy and he's going to use a lot of terms that you may not understand. Don't fret because I may not understand them as well. Ha ha. Some would say AI is the snake oil of the 21st century. It's being marketed to you in ways like never before. If you're going to go buy something because of its AI capabilities, you owe it to yourself to really know what you're being sold.  What its real capabilities are, and if it is or isn't really artificial intelligence. "

Anton Antonov:                           02:22                 "Artificial intelligence, I just don't believe in this. I think it's a great marketing term. It was used to extract lots of money from the American military machine which actually paid off  but it is a very inspiring term. Lots of people have been doing artificial intelligence in one way or the other, but there also have been to artificial intelligence winters. One was in the 60's or the second half of the 60's, and the other one, which I was actually kind of a witness, I was a teenager. It was in the 80's. And artificial intelligence basically is coming in  favour and out of favor. Right now, when they do machine learning, software tools or machine learning presentations, I see one of my goals is to speed up the coming of the next artificial intelligence winter. So, from that perspective, as I said, I don't believe in it. I believe in the so-called weak, artificial intelligence, everything is  0's and 1's."

Dave Webb:                           03:25                   "As the expert, Anton says that lots of people have extracted a lot of money from the term AI. To explore this further, we must break AI into its two main components- weak AI and strong AI. Strong AI is what you envision when someone says AI, gets the AI from the movies. Think of the Terminator, the beings with emotions who interact with us in our daily lives. That doesn't exist yet. However, weak AI does exist. This is also known as machine learning. And this is helpful for all of us. It helps automate, it helps us become more efficient. It even helps create jobs."

Dave Webb:                        04:03                 "Machine learning[inaudible] forms they are artificial intelligence. This is what we're trying to do. We're trying to solve problems which humans have been trying to solve in some mental meta mentally intensive problems using machine learning [inaudible] forms. So of course we're doing artificial intelligence. Ma Show individual  he has s a little bit more structure in.  I don't want to spread out too much in my exposition here, but basically we have the supervised machine learning, unsupervised and reinforcement learning. I would say the reinforcement learning is the closest to artificial intelligence is being, it stands for  stands for, people understand."

Anton Antonov:                   04:45            "So machine learning is our friend. All hail our robot overlords. "

 Music Sound:                   04:50             "I'm kidding of course.  Where am I? When you're applying for a job or looking for someone to fill a job, machine learning is involved in your applicant tracking system. Things like latent semantic analysis and latent semantic indexing. Turn your resume into something a machine can quickly determine is a match or not to another document. This saves recruiters time and money and it matches you with the right job. In this episode, Anton is going to talk to us about AI, what your resume looks like to the computer and what advice does a senior data scientist give to job seekers who are crafting their resumes to find their next job? From the ethics side, for the job seeker and the recruiter, does machine learning cause or prevent discrimination? Is it an advantage or is it indiscriminately filtering out the best candidate for the job ? And, can it replace the human element of the recruitment job? I would argue that it cannot.  So much to tackle in this episode of RecruiterCast.  I'm Dave Webb, CEO and co-founder of BrightMove for over 25 years. My job has been to learn all of the ins and outs of the recruiting industry and then write the software, makes it all happen.  I know who to talk to and what to ask them. We have the information that you cannot get anywhere else. So get ready to learn the secrets that will give you an advantage as a recruiter, job seeker or business owner. It's time for RecruiterCast."

Advertisement:                    06:16           "Recruiter cast is brought to you by bright move the most innovative applicant tracking system built with a recruiter and candidate in mind. Learn"

Dave Webb:                       06:28           "Today, I'm joined by Anton Antonov, who is a senior research scientist at Asindo Data. Data scientists is probably very an intimidating term to a lot of people who don't know what you do. So why don't you give us just a little bit an overview about yourself and what data science is?"

Anton Antonov:                  06:46            "Oh Yeah, this is a huge topic. So I brand myself as a senior data scientist. Right now, I say you introduced me as a research scientist, which actually I am, but I do brand myself as a senior data scientist in order to get higher paying jobs. And so, but my background is in applied mathematics, so by education trade and things I want to do, I'm an applied mathematician. And so, I started my career doing  applied mathematics that related to stuff like say a lot of skill air pollution simulations and this was the big data at the time 20 years ago. But then things evolved. Data was much more collected, curated and so forth. And right now, data science, what is dealing with, it's an amalgam of computer science tasks, data engineering, collect software engineering, data engineering tasks and statistics. It's a strong overlap. It's an overloaded term. Different people understand different things. If you say data scientist, from my perspective, data scientists, it's not my full mollation, but they're the best software engineer amongst statistician. We're the better programmers in the mathematics scientific communities and we're the better statistician or the better mathematicians when it comes to computer scientists. Does it that make sense?"

Dave Webb:                       08:10           "Yes, it does. So, what we're saying is data science is a very broad field. The amount of data we have in the world today is growing exponentially. And there's kind of like a whole new career path for mathematicians and applied math petitions to take data as we see it today, look at it as numbers and then apply algorithms to it to make decisions. And that kind of leads us and to the RecruiterCast topic of how data science and artificial intelligence kind of play into the recruiting space and specifically maybe how your resume might look to a computer and the mathematics that could be applied to that? So, I think a good place to start since this is such a broad area, I think Anton, if you could  just help clarify what is the difference between automation and machine learning and true, what we would call deep artificial intelligence. Could you just kind of explain those three things to us?"

Anton Antonov:                   09:08          "Ah  Yes. Actually not three, two of them. Artificial intelligence,  I just don't believe in this. I think it's a great marketing term. It was used to extract lots of money from the American military machine which actually paid off during the first Gulf war which they made through something called dart, like dynamic analysis and the planning tool. And I was surprised to learn that actually within four or five years of applying this tool based on a so-called artificial intelligence principles, it completely paid off the investment the American military made in that direction. But it is a very inspiring term. Lots of people have been doing artificial intelligence in one way or the other, but they also have been two artificial intelligence winters. One was in the 60's or the second half of the 60's and the other one which I was actually kind of a witness, I was a teenager. It was in the 80's. And so artificial intelligence basically is coming in favor and out of favor. Right now, when I do machine learning,  software tools or machine learning presentations, I see one of my goals is to speed up the coming of the next artificial intelligence winter. So, from that perspective, as I said, I don't believe in it. I believe in the So-called weak, artificial intelligence-everything is 0's and 1's. We basically choose to call certain methods, certain applications off mathematical algorithms to call artificial intelligence. Very often, when the application happens, people stop seeing it as artificial intelligence. Like say right now with your phone, you can just start target it to some receipt, for example, and get a recognition entity name recognition of the companies, the recognition of the sum and get the total or something, right.  Now, 40, 50 years ago, this was science fiction but right now this is artificial intelligence. People don't see it as such. They see it as an optical character recognition or whatever. And so this is like this moving past or moving goal, what exactly artificial intelligence is. Machine learning is a little bit more timid term which actually illustrates better in my opinion, what exactly is happening? I would say some people refuse to see motion and anchor already forms of artificial intelligence which I think they are wrong. Machine learning [inaudible] forms, they are artificial intelligence. This is what we're trying to do. We're trying to solve problems which humans have been trying to solve in some mental, meta mentally intensive problems using machine [inaudible] forms. So of course, we are doing artificial intelligence, machine learning though it has a little bit more structure  and, I don't want to spread out too much in my exposition here, but basically we have the supervised machine learning, supervised and reinforcement learning. I would say reinforcement learning is the closest to what artificial intelligence stands for. People understand there is some breakthroughs in neural networks, image recognition, etc. They belong to the conceptually, the simplest part of machine learning, which is supervised machine learning. Now when we do related semantic analysis, analysis of resumes and documents, we most likely engage in unsupervised machine learning. But at some point, we want to move things to be into the supervised realm for whatever kind of purposes, citizens, et cetera. The topic might be a little bit vast. But as I said, if you want to start to learn a little bit more about this, the supervised machine learning algorithm, they're conceptually busiest to understand, busiest to figure out what exactly is the methodology and busiest to kind of expand to further your knowledge about this this field."

Dave Webb:                    13:11            "I really appreciate that explanation. That was very, very helpful. So we have supervised machine learning, we have unsupervised machine learning."

Anton Antonov:                13:18            "Yeah."

Dave Webb:                    13:18            "We have what I've personally written some articles about to try to dispel this AI recruiting myth, and that's adaptive AI also known as what most people would identify as the Terminator from the Terminator movies, that does not exist. And, and we're nowhere close to that existing."

Anton Antonov:                13:38           "Right."

Dave Webb:                    13:38          "Is that correct?"

Anton Antonov:                13:39           "Yes, correct."

Dave Webb:                    13:40           "Okay. So, the idea that you're going to call a company for a job interview and the job interview is going to be conducted by a computer or an artificial intelligent being does not exist today and we're nowhere close to that?"

Anton Antonov:                 13:53           "Well, I think it can happen, but it's another thing. What is isn't going to be effective or are the outcomes of having this system in place desirable."

Dave Webb:                     14:01          "Right. What we do have in the recruiting space, and some of this has been around for a long time, at least as long as recruiting a technology has been online and computer systems and on the internet, because there wasn't a whole lot of opportunity to apply technology and mathematics when people were faxing paper resumes and they were being put into folders and then recruiters were sorting through them. But kind of with the advent of the job board and then the resume bank and then what we now have as the modern applicant tracking system, there're a lot of technology available, some of it programmed by the ATS provider, and then some of it kind of inherently built in to the database. So some of the things that we have are like text searching and beyond regular text searching, we would get into latent semantic analysis or LSA.  And it's my understanding that LSA is when you would take a head document or a bunch of documents, which I believe are called a Corpus and those documents are tokenized and then we can apply mathematical algorithms to them, not only define the occurrences of words, which would be skills that we were looking for in hiring, but maybe also the proximity of those words to other words that would indicate there's a stronger presence of that skill or more experience of that skill in the resume. Is that an accurate statement  Anton and could you kind of elaborate on LSA as it pertains to resumes and maybe even other types of documents and how that information could help us make better assessments of candidate qualifications?"
Anton Antonov:                15:40            "LSA there's first   a little bit terminology, this latent semantic analysis and latent semantic indexing. They are merely pseudonyms and I would say let's use another pseudonyms. But latent semantic indexing is a little bit harder technique, which kind of goes all the way starting from the document, the tokenization, defining of topics and making actual indexes for some changes. So this is information that retrieval or type of activity information that through all is I guess a long history at this point. It has been done in different organizations, IBM or AT&T, they started doing this. We'll latent semantic analysis is slightly newer, but latent semantic analysis officially is saying what employs the so-called distributional hypothesis, which you basically mentioned. So this is what  terms which are words which have similar semantic meanings that is  going to have similar distributions. So, if you have a sufficient amount of texts, you are going to see what this distribution of those words are similar and because of this, you can conclude with their either pseudonyms or semantic little edge in some way. Now this here, it's a very interesting relationship. Are the similarities, when you say similar distributions? What is your distribution similar location of the documents or in similar kind of neighbors of other words? You see that's actually kind of different like imagine I take some like for example, Google, before we use deep learning for a translator, translator programs, we use the 50 or what about 250 manuals translated in different languages from a UNICEF or UNESCO, one of those. My point is a large body of text translated into a lot of a number of languages. And so, what is switching the same paragraphs in just those documents, they're positionally related. They have positionally distribution so you can conclude that they have some kind of meaning.  They're in the same position in the same pages, in the same chapters or in the same first paragraph. So the first chapter, so maybe just was related. This is one way to do it. The other is to actually find synonyms, but then this means when word is presented, the synonymns of this word, they're not going to be presented. So the similarity is more, it's kind of a cross, it's not so in the position in the documents, it's not in some kind of a call service and a document but it's a repeating pattern from different types from different say sentences or paragraphs with similar structure. This tool has artificial names for official names, a symptomatic and paradigmatic in our relationships, whatever. But for latent semantic analysis, we're trying to find both distributions. Now, the most popular, I think the more powerful technique is chooses we choose as a linear vector space representations and bags of bags of words at presentations. But right, so, but this is a little bit more of a technical manner. The one with, what is in latent semantic analysis and my interest has been in it too from a mathematical perspective. When you use the letter semantic analysis model, you can actually extract topics and topics were basically related so words which are related. As I said, if I use the paragraphs, they basically appear in the same documents and paragraphs are comprised of certain subset of this topic most of the time. Like this one guy in Chicago who actually in his scientific career of 30 years, he proved that all Western books comprised of 103 ideas. So basically any kind of Western book, you pick it there. It's a sum combination, either of one or two or three of the same, no more than five ideas. That's great. So for example, typical example, matrix, you know, the matrix, it's the one, it's the same as Christ, you know, is the savior of humanity kind of like this kind of ideas. So, it’s a great book. It's fascinating.  So, basically when I started applying this to a movie recommender systems, I would know, but yeah, well most movies of course they are like, you know they will have more than say 120 topics, so I'll start, this is going to be my start and I'm going to start extracting topics for the movies and try to classify movies now, not because of the words which are in them, the bull and much, but because of the topics that are present. Imagine 2 reviewers of the movies. Talk about, say I don't know, some kind of how about mystery movie. And so, one talks about wife killing and the other one says spouse murder. Well, they're kind of the same, but there is no brilliant match. So, how do you figure out from wife killing, spouse murder? It's the same, the same phrases or the same words. This happens with topic extraction. So, similarly with resumes, when I have applied this to resumes, job descriptions, past their experiences, very often people put like they're required to put three skills in the resumes or some form they fill in and they might say iPad. Some of them might say iOS. Some of them might say escort but actually you need to figure out what all this actually, if they have programming, they are actually capable in programming Keno into the apple environment for iOS. So, making topics or for skills and finding statistical disorders for the skills. This has been extremely, extremely important task but actually improves the information retrieval when it comes to resumes and candidates. I mainly respond and we may be gang of branching out too far and too wide, but am I responding to your question?(laughs)?"

Dave Webb:                     21:44          "No, you are, that's a great answer. And I like how you kind of equate the resume topics to movie topics because everyone's familiar with movies I believe and."

Anton Antonov:                 21:52           "And Netflix presumably. Yeah. So, yeah I recommend those."

Dave Webb:                     21:55            "Yeah,  so,  I'll, I'll use some examples that came into mind as you were explaining it, things that are related but don't sound or look alike, distribution, pick, pack and ship. Those two things are very important and they're essentially the same thing. So if you're looking for someone with experience in a distribution center, they might have a pick, pack and ship or order fulfillment or a handful of other synonyms that LSA or LSI would help us recognize as synonyms and cause that resume to become a match. And that'd be makes you a more effective recruiter because you have this tool doing all this analysis on the backend. It's almost immediate to you when you're doing the Search for it."

Advertisement:                22:40             "RecruiterCast is brought to you by BrightMove, the most innovative applicant tracking system built with a recruiter and candidate in mind. Be sure to check out BrightMove on Twitter, Facebook, Instagram and LinkedIn @BrightMove. Visit us on the web at"

Dave Webb:                    22:57     	"I’m glad you brought up the nearest neighbours algorithm   because in my 20 plus years in the recruiting space, I've seen nearest neighbours use as a very effective learning tool for   new recruiters. So, you get a recruiter who's never done technical recruiting before and you say, hey, we need you to go find us a object oriented C++ programmer and the nearest neighbours can then analyse millions of resumes and say, some   other words that are really close to these are Java and C sharp and NVC or now it's agile and all of the kind of like supplementary skills that go along with coding in  a modern corporate environment in an object oriented language. So it's amazing how much information is embedded within the resume that's not even put there on purpose."

Anton Antonov:               23:48                   "Hmm, mm. Hmm, mm."

Dave Webb:                   23:49                   "That the recruiting technology can extract and use to educate recruiters and make them better recruiters. Having said that, let's talk about the resume and the resume author. There're lots of myths out there that you can google it. I've seen articles on on sites like medium and LinkedIn where the reason why you're not getting a job is because that that ATS is filtering you out. It's sorting you to the bottom and it's not your fault, it's the ATSs fault. So let's say someone believes that their resume is the problem that they're not getting callbacks. Knowing what we know about how resumes are analyzed and how recruiters are searching them, what advice would you give to someone offering their resume or revamping their resume to try to find a job to maybe help them get noticed and maybe even dispel or support the myth Anton, that."

Anton Antonov:                 24:39                "Hmm, mm."

Dave Webb:                    24: 40                "A machine can discriminate against you and your resume."

Anton Antonov:                24:43                 "This is most likely even not close to how correct this kind of impression with the system is filtering them out. Yeah, sure, but even though the system is filtering them out and putting it on the bottom. This means that they are way too many candidates. This is going to be in any kind of situation, way too much competition. Now in recruiting, this is not like say movies. I can recommend one movie to thousands of millions of people. I mean, if I am a recruiter, I cannot recommend one person to 10 companies. I was actually, so if I have enough qualified candidates for my companies I'm dealing with, I'll try to get as many candidates as possible. So if for people think what they're qualified, so, first of all, they need to decide, are they really qualified for the job that we're looking to find? Maybe they are qualified, but they demand too much higher salary or the location that they're looking at,  it's just the pool of jobs is very small. The other is that very often people would act to job if they're at job boards and et cetera. That's not necessarily a kind of honest looking for a new person for this job. This job's probably already filled by people who already by the companies who know who want to be there. They just need to go through certain process or for publishing the job on some board in order to justify whatever kind of selection they have made. They might have made the selection already. So, what I'm saying is that if you think you're discriminated, you better try to increase your network of people, word of mouth. It's a very, very good way of fighting the system if you think the system is against you. And by the way, more or less for example, this is a traditional approach at this point. It has been started like if say Obama with his with his second campaign actually popularized data science quite a lot in which instead of, gut feeling, smoking cigars in the backroom kind of guys made decisions, they actually did use, Obama did use data scientists to produce where to go next, what kind of meetings to have and, he used data science for this. So Trump basically, when he thought the system is against you, meaning the media, he actually did to an alternative approach. He used Facebook, he used Twitter and so forth. So, similarly, if we think the system is stacked against us, well, word of mouth, looking for a different job and then from this different job finding something else. Now, I personally think when the systems at play,  they're probably good enough to make the information retrieval which is needed. In most of the situations I see is that they are actually way too many candidates or way too few. There's usually not the right amount. And so this would be my kind of advice, be a little bit more scientific and shake the things up a little bit if you think you are being discriminated against. I strongly suspect that just being honest about your accomplishments and  just for all with the application process gives you more chances. I personally am very annoyed by the system which exists in place, which asks you for 20, 30 minutes to start filling out some forms. The reason for this is most of the time with the resume you have, although there's some automatic parsing systems, they're apparently not that good. I mean, I have seen this from my personal experience, and so that's why you need to kind of sit down and fill out forms. Well, now, here this when people give advice, is it a good idea to hire a consultant to do this for you? Is it a good idea to hire someone who can reshape your resume let's say for certain, I don't know, like websites from Taleo or Taleo or whatever it is. Your resume is going to be ingested much quicker and hence your application process is going to be sped up. This is the type of questions I'm going to ask. I personally, I have to say at this point or the point in which I was job hunting, the jobs I really liked and got, they were from word of mouth. And so, although I've been doing recommender systems both say in healthcare and movies and the recruiting, I do actually think the word of mouth type of recommendations in job finding is the best one. And some of the companies who are most aggressive and most interested to actually improve the hiring process. They do look into the word of mouth type of relationships and the recommenders."

Dave Webb:                        29:34                 "Yeah, that's great advice. Every job that I can remember I've gotten since in the last 25 years has been through word of mouth. Rarely has it been me applying to a job or a recruiter reaching out to me on a cold call type manner. It's always a recommendation. Touched on some really good points there, ynd some of these I have answers to and I'd like to share those with the listeners."

Anton Antonov:                     29:55               "Huh-uh"

Dave Webb:                          29.55              "If you're a recruiter and you have a candidate filling out a basically what is the equivalent of a four-page handwritten form from three decades ago online, you are losing qualified candidates because good candidates are not going to waste their time doing that all day long. If the software you're using can't take a resume and parse out basic contact information, employment history, education history, certifications, and the length of all those things, then perhaps you should look at a different solution because there are a lot of them out there that do that very well. At the same time, I've written a couple of articles about treating the job description as a marketing tool, so if you're recruiting, you really need to write your job description in a way that you attract good candidates and once you attract them, it's kind of like advertising one-0-one, you have a matter of seconds once you've peaked their interest to get them to engage with you and apply to that job. The last thing you want to do is ask them for more than a resume upload and an email, maybe even a phone number. At the same time, if you want to get in touch with that candidate, you're going to want to be texting them. It's much easier for them to respond to you via text message than a phone call, especially if they're a passive candidate currently on a job. So all of these things that Anton kind of  touched on, we've seen play out  in the real business cases that we've solved and been a part of as recruiting and recruiting providers. To that point, the thing that you said to me that really resonated, that I've always believed and always talked about is being honest on your resume, not trying to fool a system. The system is there to kind of measure honesty, measure reality, and then kind of bubble those to the top. So, if you write a really good job description, and by the way, as a recruiter, when you get a resume that's just filled with keywords, it's glaringly obvious and I promise you, your resume gets tossed. Be conversational in your job duties. Talk about what you really did, talk about team interactions, that kind of stuff. Anton, would you recommend maybe having more than one resume, depending on the type of job you're applying to because some of us who have been in the workforce for a long time, we might've started off doing one thing, but now maybe we're in management or."

Anton Antonov:                       32:10               "Right."

Dave Webb:                          32:11               " Senior roles that have a much different skill-set required than a starting role. What do you think about having multiple resumes?"

Anton Antonov:                      32:21             "This is advice which I like to give, but which I don't follow myself personally at some point, until recently actually, somebody told me, look, your resume is actually a CV. Meaning this is more of a university-based type of a resume. Maybe that's good for universities, but it's not that good for job hunting say in the U.S. And so I had to change it and also I had to shorten it much to be something much smaller. But I still do use the same resume. And there's another thing here in my conversations, I do say, look, because I'm practicing data scientist or senior data scientist, look at my get hep poster list, look at my blog posts. I don't need to prove that much to you. I mean, find lectures of me on YouTube, find some of my work on say different blogs, different articles. Others say, oh, use my software, good corp and you only figure out what exactly I'm doing. This here, it's already trying to fortify this statement that I'm a senior data scientist. Now, very early on though, before leaving actually Europe and coming to U.S, some of my professors from Academia when they moved to start working in different companies, yes, they did change their resumes. We did actually remove some of the specialties just in mathematics in order to look not necessarily more, yeah, in a sense more qualified, but they didn't spend time on some stuff which are irrelevant to the job, but they spent their career or they have spent enough time on the jobs and the tasks which are required for the institution they're applying to. And so this is basically what a personalization is and the recommenders, you trying to adapt your approach to the different companies, to the different jobs. Now, when you go back to the intellectual honesty and to then cornerstone your resume, so I'm not saying that you should change your basically change the spin of what you're doing depending on the job you're applying. In some sense, this was addressed before with cover letters as far as I understand. But, now currently, it's one thing if you're talking to millennials in Silicon Valley who have a startup, and it's another thing if you're applying to a big company in which the decision-makers follow certain process. You just need to take this into account. Having a resume which looks it's like not like a cartoon, but it has like way too many kinds of fancy kind of formatting call graphs and  etc. might not be appealing to people who are like baby boomer or so and the late stages of their career. Quite on the contrary, it might be actually found quite informative and appealing to people who grew up with the internet. So yes, of course, knowing key audience, like say what you do if you do comedy or if you do scientific presentations, it's very important advice. So similarly, knowing your audience for your resume, it's very important, and ideally, you should cater to that. "

Dave Webb:                        35:35          "You make a good point because we've established now that machines aren't making the hiring decisions. People are, machines are just tools and these algorithms are just tools to help them talk to the best people. Sometimes the best people will fall through the cracks and sometimes the worst people for the job will actually get some time but the recruiter has a limited amount of time to talk to people. So that's what we're just trying to do. We're trying to optimize and make them as efficient as possible, but there still is that human element and that's a variable. You touched on it and you made a very good point. You never know who's going to be looking at your resume, and depending on the type of job you're applying to, kind of tailoring the resume to who you think is going to be making that hiring decision is not an ethical practice at all. It's actually a strategic practice and one that is perfectly acceptable. You're still going to have to be yourself once you get the job. If you get the job, you don't want to pretend to be an outgoing person when you're not, when you have to be on a team in a room full of people with no walls for 10 hours a day for example."

Anton Antonov:                      36:40          "Right. well, in some sense you're doing a favor to all parties. If you personalize your resume or make a specific resume for a specific job. You're speeding up the process if this is really helpful. Unfortunately, you might misread the situation and the people. So then again, like this is a little bit trial and error and different people should see what works for them."

Dave Webb:                         37:04            "Okay. I'd like to change topics a little bit to something that I know is important here in the U S and that we have a whole set of laws around equal opportunity, employment and non-discrimination based on many, many factors. And these laws are always changing. States have different laws. We obviously have some federal laws that we all have to abide by. But is there a way that we could apply resume processing with machine learning to maybe prevent discrimination in the first round of evaluation?"

Anton Antonov:                     37:36            "Let me tell it the other way around, if I want to discriminate, actually, can I actually use the resume information? This is extremely difficult. If you remove like say probably names and country or origin and some other things. It's like, just from the resumes, they're usually how resumes are written. It's very hard to discriminate. It's not like say music,  like say, I mean, rap music is completely identifiable. If you do music recommender systems and you're trying to base this on lyrics, for example, only a rock to a point overlaps with the lyrics in rap. Basically Rap can be uniquely identified very quickly just by the language, for example. And country music also very uniquely identified by the dockets and discusses and et cetera. Now, in the resumes, they're just not diverse enough. I mean, I don't see how I can easily make from resumes, as I said, certain information, like a name, date of birth, date, country of origin, whatever is removed, how I can make it to successfully discriminate the candidates. So, I don't think this discrimination happens, although again, this is basically who is going to make a decision. And if the pool of candidates is large, yes, underrepresented groups are going to be underrepresented in the large pool too. So, it's not so much like discrimination is just the statistics. You have hundred candidates, but a recruiter or human resources person, they know what they need to pick up eight just purely from probability perspective. Eight candidates, just candidates, and they know they need to pick a one after that is just the probabilities get small.  I think discrimination can happen if we evaluate, say essays or writings by students I know, different type of SAT or JIRA tests or whatever is. I don't think this can be done with resumes. Resumes by the way they're written by the style and general advice, they're very technical. They're very impersonal from other perspectives. If discrimination is possible, this is mostly because like say they might be a certain correlation with say the schools who have been out or the companies who have been out or whatever. The places you have worked at, but in order to do that, you might be having such a really big pool. In order for this to be detected and to be detected and to be kind of seen as an effect, this really needs to play out on a relatively large number of people, a large number of resumes.  We make a study about this? Yeah, sure, but my problem is that usually most of the situations when you do the segmentation, location, qualifications and a certain maturity in whatever, like what kind of a career path, where in the career path people are, the candidates are not that much. And so basically it's very hard to normalize for this kind of variables and demonstrate that there's some discrimination. So again, like, do I believe that this can be done knowing everything about the candidate? Yes. Do I believe it can it be done by removing say certain like identifying information, like the things we will do, say in healthcare. We'll remove say names and we'll keep the age because that's important for healthcare, but in say we remove age, names and country of origin and some other ethnical background. No, I don't see, I don't see how the systems are going to discriminate."

Dave Webb:                       41:34                 "Yeah,  I agree with you for the most part though. The only example that I can really think of is maybe age discrimination."

Anton Antonov:                   41:41                "Yeah."

Dave Webb:                       41:41               "And that if someone's got a lot of cow ball experience and they've been doing it for 40 years, you know, they're at least 60."

Anton Antonov:                  41:46                 "Yes."

Dave Webb:                      41:46                 "Years old for example. So."

Anton Antonov:                  41:47                 "Right."

Dave Webb:                      41:48                  "I think there's a window in there where if the experience levels between five to 20 years, you wouldn't know necessarily if the person was between the ages of 30 and 50."

Anton Antonov:                  41:58                  "I like, for example, I will kind of make some interesting chronologist here I guess, but it's one of my favorite examples. Currently in data science, you either you are very senior data scientists or you are very junior. If you're intermediate level, it's like, the intermediate level data scientist as a subset of data scientists  is shrinking. It's like the middle class in U.S.  So, and one of the reasons, I mean there're several reasons for this, but basically one of these is things to automate is machine learning and data science workflows. One of the easiest things to automate with machine learning is the automation of machine learning and data science workflows. Because of this, many companies think, Oh, I don't need to have that many senior people. I need to hire one senior people and bunch of juniors and I can care for a good term performance team. So why I'm telling all this, I personally in interviews I conduct, I age discriminate, I discriminate young people. I discriminate old people. For me, it's really very important to know how my co-workers or future coworkers are going to be thinking where they come from. And say data science relatively it's not a new term, it's a very old term but as a profession, being advertised, it just came relatively recently into play. And because of this, all the data scientists, they usually have been doing something else. And so they had been doing some, if they have done engineering in physics, I actually would like to talk to them. Now, the young data scientists, they have graduated from say different schools like say Florida The Atlantic University or Georgia Tech exactly as data scientists.  What exactly data science is?They have been taught by people who are not data scientists. They're probably statistician teaching data science or computer science professors teaching data science. So what exactly these people are doing? They have completed the status of a data science but what does this mean? So that's what I'm saying. I will be discriminating because of this because the age is a very strong indicator of both educational background but also motivational backgrounds. People who can do re-qualify from engineering core physics, all the people into data science, they're different from people who decided to be data scientists because there's an explosion of the data scientist jobs. It's just a different level of motivation, different level, different reasons for acquiring knowledge. Does this make sense?"

Dave Webb:                       44:22                 "Of course it does."

Advertisement:                  44:24                 "Be sure to check out RecruiterCast on Twitter, Facebook, Instagram and LinkedIn @RecruiterCast."

Dave Webb:                       44:31                 "What inspired the data scientists of the 60's or the 80's or 2000's is different than what inspiring them today. Computer science kind of went through the same transformation."

Anton Antonov:                   44:41              "Correct, yes"

Dave Webb:                      44:42                "Right, in  the late 90's.

Anton Antonov:                  44:44                "Gold rush."

Dave Webb:                      44:45               " I'll just take."

Anton Antonov:                  44:46                "Gold rush in a computer sense."

Dave Webb:                       44:46               "Well, yeah, it was a gold rush, but it was kind of like the doctors and lawyers of the 80's it's like whatever is on the USA today top 10 best paying degrees lists."

Anton Antonov:                   44:55              "Yes."

 Dave Webb:                      44:45              "Sees this urge and that's just reactionary behavior, yeah.

Anton Antonov:                    44:49              "Yes, correct, yes."

Dave Webb:                        45:00              "We would expect with anything."

 Anton Antonov:                    45:01             "So right, because of this, yeah, age is discrimination. For me, when I say discrimination depends how we define it, but I would basically do segmentation, age segmentation. Now, in each different age segment, I'll discriminate differently. Let me give it more detail and be more clear about when I say age discrimination. No, I don't do age discrimination. I do age segmentation and then I do different discrimination strategists within each segment.  Does it makes sense now?"

Dave Webb:                        45:32              "It makes sense and I think we should point out that discrimination is an actual mathematical term. It's not necessarily synonymous with the act of giving someone  a bad hand based on."

Anton Antonov:                    45:32             "And not, yes, and I do."

Dave Webb:                        45:47             "Something I just want to clarify that."

Anton Antonov:                    45:49            "Right. I do actually. I don't discriminate just because of some inherent properties of the applicants. I'm more like trying to evaluate motivation and background. Now, to be honest, sometimes I'll start teams of people who think too much like me, although I would welcome them, but I don't necessarily look for people who think like me for example, of people who are equally good. And also, I don't look for people. I don't want to be the smartest person in the company also. So, actually I like to kind of meet people and work with people who are smarter and more capable than me. But also sometimes people, especially in say, mathematics, data science, scientific disciplines, they're looking for certain type of peers and they're looking for certain type of kind of equality, not equality, but similar fashion of fault to be followed. And no  I think that's myopic and we need to have diversity of fault, but I am going to be again, I am going to try to ask quite a lot of hard questions just in order to see how people are going to react to them. In many of the situations I work with, the problems are open-ended, the problems also the business goals change not necessarily very dynamically but quickly enough. So actually being pragmatic and not just following some procedures which they've learned in school. It's very important for me. I do actually want to work with people who can both see the big picture, but because they know the big picture, they're very good at addressing the details in the job."

Dave Webb:                         47:33            "And I think that's a great example of why you should just be authentic and be yourself on your resume because you can't predict who the person hiring is looking for. You can't assume they're looking for a carbon copy of themselves and for long-term happiness in any role, you're going to want to be doing something that you love and you're going to want to do it in the way that's always made you happy doing it because hopefully, that's why you chose that career. "

Music playing                     47:57

Dave Webb:                        47:59               "Okay, just to summarize here, the best way for the computer, the applicant tracking system and the machine learning to find you is for you to authentically represent who you are and what you've done and what it is that you would like to do on your resume. It's okay to have more than one resume and market your skills to a specific job, especially if you've been in an industry for 20 or 30 years and there're all kinds of job functions that you can do and want to do.  It will really matter in getting you found by the right recruiter for the right job. And when it comes to AI, just remember AI's great at movies. It's science fiction. Hopefully one day it'll be true. There's a lot of smart people working to make that happen, but at least today and in recruiting software, AI is narrow. AI is machine learning. It's automation, it's using really complicated math algorithms and large datasets to pretty accurately predict what may or may not happen based on the past. So, that's a great tool to have in your applicant tracking system and use it to help find people for jobs like someone you're already looking at, or use it to filter out a thousand applicants that really aren't good for the job so you can focus on the 10 that might be."

Music Playing                    49:07

Dave Webb:                       49:07                  " Can you tell me what was your most awkward job interview and what made it awkward?"

Anton Antonov:                   49:12                 "They have not been awkward in a sense like awkward interviews is like some misalignment of expectations, which actually does happen because the recruiter is sometimes being aggressive or not understanding the qualifications, they would match me up with companies or people and I'm not the guy they're looking for. I would say the first company I came here to the US, it's called Wolfram Research. They make a toolbox for mathematical computations called Mathematica or there's a engine computational search engine called Wolfram Alpha. Anyway, their info, very famous company, used a lot in academia, at least at the time, the first interview I had with them, it was complete.  I don't know, ultimately would have been embarrassing what I know now about the company, but I was a PhD student in Denmark and I found that because we're such an academic company, very close to the Illinois University, but they were just looking for fresh blood from the Academia in order to get to know what the academia is doing. I did not realize what the internships at that company they're actually job interviews and potential future employment. This was a pleasant surprise, I have to say. So when it happened, I actually did decide to leave Denmark. I was doing my Ph.D. in Denmark to join this company in Illinois. Later on, because of the word of mouth and I wanted to get my Ph.D. studies covered, I would say I came and got about many awkward job interviews. But it has been the other way around. I'm a little bit surprised also when unqualified candidate, like when I have been interviewed in this situation on both sides, it has been somewhat I mean now it's amusing, but it's like for five minutes we're talking about stuff and then realizing actually we're talking about completely different things. Whatever the expectations about the job in the interview are, we act very differently, and so that's why right now I always ask people if I'm banking interviewed, did you read my resume? If I'm going to interview someone, I do send them my resume and my background. I do want everything to be as much clear as possible before the interview."

Dave Webb:                      51:38            "Pick the top three pieces of advice that you could give either recruiters and doing resume searching or job seekers crafting their resumes, what do you think are the main top three takeaways that we could that we could listen to with regards to resume processing and LSA and LSI?"

Anton Antonov:                  51:57           "Yeah, okay, you clarified but in regards to those, otherwise, my advice would have been in some other things and in some other areas of human activities. Well, I believe in the law of large numbers but keep in mind that I'm an European who moved to U.S, and so this means that actually, in some sense I did believe during the cold war about how capitalism was advertised in the U.S. I was surprised to find that it's not so much of a wild capitalism, but one of the big things about the U.S is that for whatever reasons both companies and employees want to be fungible or let me rephrase this. Companies want to hire and fire people easily and people want to be able to move around easily. This actually is truer, the second part, people want to move isn't around. This is more like for the more higher-paying jobs or more like non-blue collar jobs, but this kind of thing you should leverage that. You should leverage the fact that it's very, in  U.S, people want to kind of hire and fire at will.  In Europe for example, that brings to quite a lot of stagnation. It's very hard to hire people and it's also in most of the Western countries I'm talking about and it's also very hard to fire them. Here in the U.S, especially if you want to find a job which really fits your personality or things you want to do. I mean, U.S is like that. Unfortunately, because of a lot of outsourcing and et cetera, This pool changed, got skewed quite a lot  and right now there's a strong, drive for a so-called gig economy. Now, I'm not, I'm against it. It seems to me that and also I lived in socialism, so I believed in some of ideas of the socialism, but you see in the 60's or 50's, there's some kind of very violent even bloody kind of encounters between the employers and the workers in order to get certain union rights and certain rights about the employee in the U.S and now with the gig economy, this is basically circumvented all this kind of rights being fought, people here fought about being circumvented but with the gig economy. Well, find another gig or two or five. So, this is actually one of the things which I would say I would like in U.S, if you're proactive, you always can find a way to diversify your options. And this is related to this leveraging capitalism means, it's the second advice, especially if you're,  it doesn't matter how young or old you are, try to diversify your skills. Ideally they need to be complimentary. If you have time and ability, this is the thing I would say like for example I mean, I don't know. You can be I know an example about you can be an architect, but you also can be a professional fisherman-someone who can actually go to the fish and fish boats and thrillers and do and do the job. And so, also, similar like it happens very often, if you're a broker or maybe a banker, yoga instructor, it's a way to diversify your kind of abilities to earn and find jobs. And I want to say the third thing, I don't know, like this was some of the anecdotes we have been hearing in Europe about America. You don't get rich by working hard. So, I mean, people who get rich, they're usually for investment or for money. Even if you look into in U S  now, it's like the entrepreneurs, the people with ideas, they're not necessarily the ones who get rich from their ideas, it's actually the venture capitalists who get rich and it's not the writers and the authors, it's actually the publishers who get rich and so forth. And you can say the same in music too. So, well try to use that and try to invest.  There's a way, I mean I've been talking to younger people here in U.S to start to invest from very early on. This actually for 1 K and etc. and all like this kind of stuff like using investments. This is something I started much later in life to be able to invest in that way. And so here in U S, We're using key investments using an alternative way of producing income probably later in life but being consistent about it I think is very important and it's one of the great opportunities or one of the great things provided by U.S as a country."

Dave Webb:                    57:06       "If you don't mind, if someone wants to learn more about Asindo or yourself personally, what's the best way to do that?"

Anton Antonov:                57:14        "They can type Mathematica for prediction and they'll find my blogs and get posters."

Music                         57:21

Dave Webb:                    57:25          "That does it for this episode of RecruiterCast. My brain hurts, but in a good way, the best way possible actually. I feel like I've learned so much from Anton and I could talk to him for hours and I think that he could talk to me for hours too. Now, that's a lot to digest, right? Did you think we are going to be talking about the Terminator the whole time? Should I make an 'I'll be back' joke right now? A huge thanks to Anton Antonov for talking with us today. That was quite an episode. Remember to hit us up on Twitter, Instagram, LinkedIn and Facebook @ RecruiterCast. Head on over to the website, and submit some questions. We have our Q and A episode coming up soon. Hey, it might even be live. We might even call you if you can take phone calls at work. You can request guests. Tell us how your day's doing. I'd also like to know if you have a dog and what your dog's name is, the breed, and send us a picture. We love animals, so let's hear from you. You can even call us."

Female speaker:                58:19               "(904) 525 8134."

Dave Webb:                     58:25              "Thanks for listening and as always, happy recruiting from your RecruiterCast host, Dave Webb, St. Augustine Beach, Florida, out. RecruiterCast is an original production produced and recorded in St. Augustine Beach, Florida, and is hosted by me, Dave Webb. Our executive producers are Andrew Seward and Heidi green, original music by Dave Webb and Andrew Seward."


Thank you for your interest in RecruiterCast. To best route your inquiry, please select one of the following:

Guest RequestSponsorship InformationEpisode Idea
Follow Us
Call Us