WEBVTT

00:00:08.280 --> 00:00:10.140
Hey everyone, welcome back to the Data and AI

00:00:10.140 --> 00:00:13.039
with Mukundan podcast. Today's episode, it's

00:00:13.039 --> 00:00:16.600
close to home. It's about rejection. Not the

00:00:16.600 --> 00:00:18.780
romantic kind. Not even the friendzone kind.

00:00:19.500 --> 00:00:23.440
It's about that cold robotic kind. The one that

00:00:23.440 --> 00:00:28.679
slides into your inbox uninvited. It says...

00:00:30.600 --> 00:00:33.359
Thank you for your interest in the role at Company

00:00:33.359 --> 00:00:36.500
X. After careful consideration, you've decided

00:00:36.500 --> 00:00:39.679
to move forward with other candidates. Yeah,

00:00:39.759 --> 00:00:42.920
we've all seen that. And it sucks. What if I

00:00:42.920 --> 00:00:45.740
told you that in many cases, it wasn't even a

00:00:45.740 --> 00:00:48.600
person who rejected you? And I'm sure by now

00:00:48.600 --> 00:00:52.659
you know how rejections work in companies. But

00:00:52.659 --> 00:00:56.899
I just wanted to give you a deeper dive of how

00:00:56.899 --> 00:01:07.159
an AI -powered ATS... Resume Checker works. Today,

00:01:07.400 --> 00:01:10.799
I'm telling you the story of how I built an AI

00:01:10.799 --> 00:01:14.500
-powered ETS resume checker, why I built it,

00:01:14.540 --> 00:01:18.000
how it failed my own resume, and what I learned

00:01:18.000 --> 00:01:20.859
that changed everything. Stay with me because

00:01:20.859 --> 00:01:23.140
if you've ever submitted a resume and heard nothing

00:01:23.140 --> 00:01:27.000
back, this episode is for you. Let's talk about

00:01:27.000 --> 00:01:30.510
the pain of rejection. It's easy to brush it

00:01:30.510 --> 00:01:32.790
off when you're told, yeah, it's not personal.

00:01:33.090 --> 00:01:36.950
But you and I both know, it feels personal. Especially

00:01:36.950 --> 00:01:40.310
when you spend hours tailoring your resume. Editing,

00:01:40.409 --> 00:01:44.810
rewriting, rewording, optimizing, reading the

00:01:44.810 --> 00:01:48.170
job description 20 times. Even using those resume

00:01:48.170 --> 00:01:53.090
scan tools. And still, nothing. For us the most,

00:01:53.250 --> 00:01:57.489
it's just that it's not just the no. It's the

00:01:57.489 --> 00:02:01.150
silence. It's the lack of explanation, the mystery.

00:02:02.390 --> 00:02:06.590
Did someone read my resume? Did it get lost?

00:02:07.530 --> 00:02:12.689
Was it filtered out? Did I get ghosted by an

00:02:12.689 --> 00:02:17.129
algorithm? Now that one question, the last one,

00:02:17.150 --> 00:02:21.409
kept me up. And one night, I just decided to

00:02:21.409 --> 00:02:24.129
stop wondering. I decided to actually investigate.

00:02:27.500 --> 00:02:31.699
What even is an ATS? Now let's break down the

00:02:31.699 --> 00:02:35.060
villain of this story, the ATS. The ATS stands

00:02:35.060 --> 00:02:38.259
for Applicant Tracking System. And it's not a

00:02:38.259 --> 00:02:42.960
new thing. It's been there for a while now. But

00:02:42.960 --> 00:02:45.680
it's gotten more powerful, more widely used,

00:02:45.840 --> 00:02:49.780
and more automatic than ever before. And here's

00:02:49.780 --> 00:02:53.240
how it works. Now imagine a digital gatekeeper.

00:02:54.439 --> 00:02:57.699
It doesn't care how hard you've worked. It doesn't

00:02:57.699 --> 00:03:00.020
care about your GPA or how many side projects

00:03:00.020 --> 00:03:02.780
you've done. It scans your resume for structure,

00:03:02.979 --> 00:03:07.219
keywords, format, alignment. It reads it like

00:03:07.219 --> 00:03:11.060
a robot. Because it is a robot. Your bullet points,

00:03:11.300 --> 00:03:15.039
if not phrased the right way, may never trigger

00:03:15.039 --> 00:03:18.240
its internal scoring system. Now, if your file

00:03:18.240 --> 00:03:21.740
isn't readable, if your text is embedded in a

00:03:21.740 --> 00:03:25.120
weird format, if your font is non -standard,

00:03:26.030 --> 00:03:29.150
I mean, just toss your resume aside. And the

00:03:29.150 --> 00:03:31.330
worst part? You'll never know what happened.

00:03:33.610 --> 00:03:36.189
So, you know, one of these nights I did hit a

00:03:36.189 --> 00:03:38.530
wall. And like I said, I did hit a wall on that

00:03:38.530 --> 00:03:43.009
particular night. It was after getting these

00:03:43.009 --> 00:03:46.610
rejection mails time and again. And I remember

00:03:46.610 --> 00:03:49.330
one in particular. I'd done everything right.

00:03:49.750 --> 00:03:53.879
The resume looked great. Or in my opinion, of

00:03:53.879 --> 00:03:56.340
course. But then clear formatting, tailored language,

00:03:56.659 --> 00:04:00.740
keywords from the job post, and then came the

00:04:00.740 --> 00:04:04.280
gold robotic rejection. Something in me snapped.

00:04:05.659 --> 00:04:09.500
I opened my laptop, spun up a new folder. I said,

00:04:09.580 --> 00:04:14.159
all right, let's just build this thing. My goal

00:04:14.159 --> 00:04:19.720
was to build a tool that mimics the ETS. Not

00:04:19.720 --> 00:04:23.899
just any checker. a tool that thinks like the

00:04:23.899 --> 00:04:29.220
robot. So here's what I used to do this. Python,

00:04:29.439 --> 00:04:32.939
the programming language of my choice. Streamlit,

00:04:33.120 --> 00:04:36.480
an amazing Python -based library for the front

00:04:36.480 --> 00:04:39.759
end. So to have like a cool user interface and

00:04:39.759 --> 00:04:44.259
very easy to build one, Streamlit is a great

00:04:44.259 --> 00:04:49.319
option to do this. OpenAI for the natural language

00:04:49.319 --> 00:04:54.389
insights. scikit -learn for similarity scoring

00:04:54.389 --> 00:04:58.189
so similar scikit -learn is if you're not familiar

00:04:58.189 --> 00:05:03.709
already has been used widely for machine learning

00:05:03.709 --> 00:05:08.009
applications in python so that is a great python

00:05:08.009 --> 00:05:16.350
based library then pypdf2 and docx2 text so pypdf2

00:05:16.350 --> 00:05:26.389
is pypdf2 and docx to text is d o c x 2 t x t

00:05:26.389 --> 00:05:31.389
for file parsing and i wanted to recreate the

00:05:31.389 --> 00:05:36.189
gatekeeper this digital gatekeeper but also add

00:05:36.189 --> 00:05:41.850
something it doesn't have feedback so the ats

00:05:41.850 --> 00:05:44.389
rejects you but it doesn't really tell you anything

00:05:44.389 --> 00:05:48.290
specific and That's why I wanted to build that

00:05:48.290 --> 00:05:52.089
feedback process in. I didn't want to just check

00:05:52.089 --> 00:05:55.810
a resume. I wanted to explain what's missing,

00:05:55.930 --> 00:06:01.889
offer actionable improvements, give people a

00:06:01.889 --> 00:06:07.029
fighting chance. And then I tested it on my own

00:06:07.029 --> 00:06:11.829
resume. I uploaded my resume, pasted the job

00:06:11.829 --> 00:06:17.790
description, and clicked run. The result? A cosine

00:06:17.790 --> 00:06:22.610
similarity of 43 .43%. A GPT -4 score of 85.

00:06:23.370 --> 00:06:29.829
Verdict. Fail. Cosine similarity is just a similarity

00:06:29.829 --> 00:06:33.610
score of my resume and the job description. Right?

00:06:35.870 --> 00:06:39.129
Just seeing how far off it was. And GPT -4 score

00:06:39.129 --> 00:06:42.110
is nothing but the scoring system from OpenAI.

00:06:43.029 --> 00:06:47.620
GPT -4 is the GPT being used by OpenAI. And verdict

00:06:47.620 --> 00:06:51.259
was failed. Yep, my own resume that I'd used

00:06:51.259 --> 00:06:58.079
multiple times for many a jobs had failed. Now,

00:06:58.139 --> 00:07:00.420
not because that I didn't have the experience,

00:07:00.639 --> 00:07:05.000
but because the ATS didn't just see it that way.

00:07:05.899 --> 00:07:09.800
It flagged missing keywords. It said Google Analytics

00:07:09.800 --> 00:07:14.079
4, retail, site measurement. All those were missing.

00:07:15.060 --> 00:07:17.660
I had the experience. I just didn't speak the

00:07:17.660 --> 00:07:23.199
language. I hadn't translated my story into machine

00:07:23.199 --> 00:07:27.000
readable success. And that was the aha moment.

00:07:30.160 --> 00:07:35.600
So here's what I learned. Lesson one, don't assume

00:07:35.600 --> 00:07:39.680
your resume speaks for itself. I thought my resume

00:07:39.680 --> 00:07:43.740
was great. Environment not getting these interview

00:07:43.740 --> 00:07:48.329
calls. But it wasn't that. i just assumed wrong

00:07:48.329 --> 00:07:52.569
that would be my take from this is i wouldn't

00:07:52.569 --> 00:07:55.870
assume that my resume speaks for itself lesson

00:07:55.870 --> 00:08:00.069
two keyword alignment matters more than clever

00:08:00.069 --> 00:08:04.930
phrasing so you're not just talking about the

00:08:04.930 --> 00:08:07.110
keywords that you have in your resume but you

00:08:07.110 --> 00:08:09.350
also need to align it to the job description

00:08:09.350 --> 00:08:17.529
right lesson three is Formatting structure and

00:08:17.529 --> 00:08:21.490
even word order changes how an ATS scores you.

00:08:22.069 --> 00:08:24.990
So whenever you're writing a resume, you need

00:08:24.990 --> 00:08:27.889
to pay attention to your formatting, to your

00:08:27.889 --> 00:08:30.269
structure, is your resume flowing through smoothly,

00:08:30.370 --> 00:08:33.450
and even how your words are flowing in order.

00:08:35.950 --> 00:08:40.629
I realized I needed to be more intentional. I

00:08:40.629 --> 00:08:43.990
swapped collaborating with product teams to Led

00:08:43.990 --> 00:08:46.929
EB tests with product and marketing teams using

00:08:46.929 --> 00:08:50.669
SQL and BigQuery. I had done this. It's just

00:08:50.669 --> 00:08:53.389
something that wasn't getting translated correctly.

00:08:54.309 --> 00:09:01.509
Right? I explicitly listed GA4, which was missing.

00:09:01.629 --> 00:09:05.389
And I added quantifiable wins. I tailored it

00:09:05.389 --> 00:09:08.610
tightly to the job. Yeah, I'm sure by now, if

00:09:08.610 --> 00:09:10.610
you've worked on your resume ever, you would

00:09:10.610 --> 00:09:13.700
know that they want a more quantifiable... And

00:09:13.700 --> 00:09:17.779
maybe a couple of occasions where it was missing

00:09:17.779 --> 00:09:24.820
for me. But I think what I changed was I tailored

00:09:24.820 --> 00:09:30.259
those quantifiable wins more to the job. Right?

00:09:30.620 --> 00:09:34.940
More customized to the job. And then I ran it

00:09:34.940 --> 00:09:39.879
again. Pass. That was the result. Same story,

00:09:40.120 --> 00:09:46.950
told differently. So it wasn't the rejection

00:09:46.950 --> 00:09:51.649
that broke me. This is... It was the feeling

00:09:51.649 --> 00:09:58.409
of being invisible. That my work, my sweat, my

00:09:58.409 --> 00:10:01.470
2am sprint to hit that deadline meant nothing

00:10:01.470 --> 00:10:05.149
to a machine. But here's the truth. If you're

00:10:05.149 --> 00:10:07.129
not tuning your resume to pass the gatekeeper,

00:10:07.289 --> 00:10:11.549
you may never reach human eyes. That's not fair,

00:10:11.649 --> 00:10:16.570
but it is real. You don't need to rewrite your

00:10:16.570 --> 00:10:20.490
life story. You just need to translate it. Translate

00:10:20.490 --> 00:10:23.230
it for the machine. So the human on the other

00:10:23.230 --> 00:10:28.590
side can finally see you. And that's what we

00:10:28.590 --> 00:10:30.009
aim to achieve when we are sending our resume

00:10:30.009 --> 00:10:32.669
across. And I didn't want to make it as a story,

00:10:32.750 --> 00:10:36.009
but this is how it is, right? And if this story

00:10:36.009 --> 00:10:39.289
resonated with you, try the tool. And it is linked

00:10:39.289 --> 00:10:42.110
in the show notes. Upload your resume, add a

00:10:42.110 --> 00:10:46.240
job description, get your score. And make changes

00:10:46.240 --> 00:10:51.820
accordingly. You don't need to start over. You

00:10:51.820 --> 00:10:56.759
need to just tune the signal. Well, thanks for

00:10:56.759 --> 00:10:58.679
listening to another episode of Data and AI with

00:10:58.679 --> 00:11:01.799
Mukundan. If you like the story, share it, subscribe,

00:11:02.159 --> 00:11:04.600
drop a review. I'll be back soon with another

00:11:04.600 --> 00:11:08.580
real -world AI journey. Until then, stay curious

00:11:08.580 --> 00:11:09.879
and stay seen.
