WEBVTT 00:00:06.190 --> 00:00:10.669 It was 2 a .m. again. I had 10 tabs open. LinkedIn, 00:00:10.970 --> 00:00:14.410 Glassdoor, my resume doc, a spreadsheet of roles 00:00:14.410 --> 00:00:17.570 I was tracking, and my inbox full of, we regret 00:00:17.570 --> 00:00:20.429 to inform your messages. And the glow of my screen 00:00:20.429 --> 00:00:23.670 was the only light in the room. I remember thinking, 00:00:23.890 --> 00:00:28.190 how is job hunting this, this, this hard, right? 00:00:29.429 --> 00:00:32.590 I work with data for a living. I automate everything. 00:00:33.200 --> 00:00:36.859 Did dashboards, reports, metrics, insights. But 00:00:36.859 --> 00:00:39.960 here I was doing the most manual process of all. 00:00:40.340 --> 00:00:44.079 Well, trying to get a job. And it wasn't just 00:00:44.079 --> 00:00:49.100 about rejection. It was about inefficiency. Well, 00:00:49.200 --> 00:00:53.159 every role that I've applied to, that I'm seeking 00:00:53.159 --> 00:00:56.380 to apply to as well, they have a slightly different 00:00:56.380 --> 00:01:01.119 task. Like every recruiter wants a slightly different 00:01:01.119 --> 00:01:05.969 version of me. And I'm sure if you have applied 00:01:05.969 --> 00:01:11.069 to jobs, you know what to do, right? Basically, 00:01:11.069 --> 00:01:14.890 you have to tweak your resume for every role 00:01:14.890 --> 00:01:17.989 that you're applying to. I realized that if I 00:01:17.989 --> 00:01:20.370 wanted to survive this mentally and emotionally, 00:01:20.650 --> 00:01:24.930 I needed a system. And since I built AI tools 00:01:24.930 --> 00:01:28.450 for a living, well, I decided to build one for 00:01:28.450 --> 00:01:32.769 myself. And that's how this all began. My experiment 00:01:32.769 --> 00:01:39.189 to fix my job hunt with AI agents. Hey, I'm Mukundan 00:01:39.189 --> 00:01:41.189 and this is Data and AI with Mukundan, the show 00:01:41.189 --> 00:01:43.890 where we learn AI by building. In this episode, 00:01:43.950 --> 00:01:46.349 I'm going to tell you the full story of how I 00:01:46.349 --> 00:01:50.370 used three AI agents, my researcher, writer and 00:01:50.370 --> 00:01:53.030 reviewer, to completely transform the way I applied 00:01:53.030 --> 00:01:55.730 for jobs. We'll go behind the scenes into how 00:01:55.730 --> 00:01:57.489 I built them, how they talked to each other and 00:01:57.489 --> 00:01:59.409 how they saved me hours of stress every week. 00:02:00.010 --> 00:02:02.030 But more importantly, I'll share how this process 00:02:02.030 --> 00:02:04.790 reshaped my confidence, my clarity, and my career 00:02:04.790 --> 00:02:07.790 direction. Because this isn't just some coding 00:02:07.790 --> 00:02:09.930 project that I'm trying to do. It's a lesson 00:02:09.930 --> 00:02:12.449 in self -automation, the kind of system that 00:02:12.449 --> 00:02:17.349 saves us not just time, but also gives us sanity. 00:02:18.169 --> 00:02:22.110 Well, at least it did for me. So, well, if you 00:02:22.110 --> 00:02:23.789 ever look for a job, you know this feeling, right? 00:02:24.050 --> 00:02:27.490 You wake up, you open whatever job site, say 00:02:27.490 --> 00:02:30.860 LinkedIn. And suddenly your day disappears. You're 00:02:30.860 --> 00:02:33.000 comparing roles, copy pasting bullet points, 00:02:33.300 --> 00:02:37.000 adjusting titles, tweaking phrases like data 00:02:37.000 --> 00:02:40.860 -driven insights for the umpteenth time. And 00:02:40.860 --> 00:02:45.199 I was in that loop for months. Felt like it was 00:02:45.199 --> 00:02:50.060 just tiring. It was just completely disheartening. 00:02:50.759 --> 00:02:53.120 There were days I'd spend three hours on a cover 00:02:53.120 --> 00:02:57.340 letter. And it's not just because I'm just sitting 00:02:57.340 --> 00:02:59.780 and thinking of what to write, but mostly because 00:02:59.780 --> 00:03:02.500 I'm procrastinating as well. Only to realize 00:03:02.500 --> 00:03:05.759 also the application portal wanted a 250 character 00:03:05.759 --> 00:03:09.319 summary instead. Or was the system a timeout 00:03:09.319 --> 00:03:15.120 and everything I wrote was gone? Now, one night 00:03:15.120 --> 00:03:17.520 I remember staring at a job description that 00:03:17.520 --> 00:03:20.300 said, we're looking for a self -starter who can 00:03:20.300 --> 00:03:23.379 work cross -functionally, leverage data to drive 00:03:23.379 --> 00:03:27.389 growth and... communicate insights clearly i 00:03:27.389 --> 00:03:29.949 thought that's literally every job description 00:03:29.949 --> 00:03:33.629 and that's when the thought hit me why am i doing 00:03:33.629 --> 00:03:37.509 this manually when i could use ai to do the repetitive 00:03:37.509 --> 00:03:40.849 parts if i could automate data cleaning why not 00:03:40.849 --> 00:03:45.110 automate resume tailoring if gpt can analyze 00:03:45.110 --> 00:03:49.050 text i can't analyze job descriptions too and 00:03:49.050 --> 00:03:52.110 the next day i decided to just build my own job 00:03:52.110 --> 00:03:56.780 hunt system with Three roles. Meet these roles 00:03:56.780 --> 00:04:00.379 now. Meet the agents whom I've assigned a role 00:04:00.379 --> 00:04:04.520 for each of these agents. The first agent is 00:04:04.520 --> 00:04:08.139 the researcher who reads the job post, extracts 00:04:08.139 --> 00:04:12.280 the requirements and highlights key skills. The 00:04:12.280 --> 00:04:16.420 writer agent uses my actual career history to 00:04:16.420 --> 00:04:19.600 tailor a resume and a cover letter for that specific 00:04:19.600 --> 00:04:24.360 job. The reviewer. It critiques both for tone, 00:04:24.459 --> 00:04:29.920 clarity, and ATS readability. Now, together they'd 00:04:29.920 --> 00:04:33.560 become my little AI team. Let me walk you through 00:04:33.560 --> 00:04:36.660 how this works. First, I paste the job description 00:04:36.660 --> 00:04:40.279 into my researcher agent. It scans it for signals, 00:04:40.459 --> 00:04:45.399 not just keywords, but intent. For example, if 00:04:45.399 --> 00:04:48.100 the posting says cross -functional collaboration, 00:04:48.560 --> 00:04:52.480 five times. The agent infers that communication 00:04:52.480 --> 00:04:57.120 and teamwork are key. If it mentions SQL, Tableau 00:04:57.120 --> 00:05:01.959 or Looker, it classifies the stack. Now it outputs 00:05:01.959 --> 00:05:04.699 a clean summary that looks like this. This role 00:05:04.699 --> 00:05:07.959 focuses on driving business insights using SQL 00:05:07.959 --> 00:05:11.759 and visualization tools. Collaboration and stakeholder 00:05:11.759 --> 00:05:15.500 storytelling are central. The company emphasizes 00:05:15.500 --> 00:05:19.420 experimentation and product growth. That's my 00:05:19.420 --> 00:05:22.720 researcher agent doing its job, cutting through 00:05:22.720 --> 00:05:26.759 the noise. And then the writer agent steps in, 00:05:26.819 --> 00:05:30.500 takes my resume data, stored in structured JSON, 00:05:30.759 --> 00:05:34.860 like job titles, accomplishments, metrics, and 00:05:34.860 --> 00:05:37.720 rewrites key sections to match the tone and the 00:05:37.720 --> 00:05:41.699 focus of the job. So instead of saying, let data 00:05:41.699 --> 00:05:45.319 analysis for marketing campaigns. It becomes 00:05:45.319 --> 00:05:47.839 collaborated with cross -functional teams to 00:05:47.839 --> 00:05:51.279 drive experimentation insights, improving campaign 00:05:51.279 --> 00:05:55.500 ROI by 23%. It's subtle, but it speaks the language 00:05:55.500 --> 00:05:58.180 of the recruiter. Now, finally, the reviewer 00:05:58.180 --> 00:06:01.040 agent takes both the job description and the 00:06:01.040 --> 00:06:03.899 writer's output and scores everything for clarity, 00:06:04.040 --> 00:06:07.819 impact, and ATS compatibility. It flags cliches. 00:06:08.819 --> 00:06:12.480 It rewrites passive sentences, where it even 00:06:12.480 --> 00:06:16.079 checks your formatting length. Now in 90 seconds, 00:06:16.279 --> 00:06:21.180 I had what used to take me 90 minutes. So let's 00:06:21.180 --> 00:06:24.220 look at part 3 where I'm turning this process 00:06:24.220 --> 00:06:27.980 into a product. Once I had the agents working 00:06:27.980 --> 00:06:31.319 individually, I wanted them to work together. 00:06:31.819 --> 00:06:34.839 I wrote a small orchestration script, kind of 00:06:34.839 --> 00:06:37.620 like a relay race. So when I paste a new job, 00:06:37.759 --> 00:06:40.980 it triggers the researcher agent. Once the summary 00:06:40.980 --> 00:06:44.779 is ready, it hands that off to the writer. When 00:06:44.779 --> 00:06:46.959 the writer is done, the reviewer takes over. 00:06:47.519 --> 00:06:51.579 At the end, I get three files. A tailored resume, 00:06:51.819 --> 00:06:55.399 a personalized cover letter, and a short LinkedIn 00:06:55.399 --> 00:06:59.740 summary. All ready to review. Or it was just 00:06:59.740 --> 00:07:03.540 magic. But more than that, it changed how I approached 00:07:03.540 --> 00:07:08.420 the job search emotionally. Before, every rejection 00:07:08.420 --> 00:07:12.560 felt personal. Every application felt heavy. 00:07:14.199 --> 00:07:19.060 It felt like data, like a test. Each rejection 00:07:19.060 --> 00:07:24.240 became a new iteration for the AI. Now, I started 00:07:24.240 --> 00:07:27.160 tracking patterns, which versions got callbacks, 00:07:27.300 --> 00:07:31.199 which tone landed better, which skills got flagged 00:07:31.199 --> 00:07:34.459 more often. Slowly, my AI system was learning 00:07:34.459 --> 00:07:40.319 me. Now, let's learn about the psychology of 00:07:40.319 --> 00:07:43.670 rejection. Let's take a pause as well. Because 00:07:43.670 --> 00:07:47.209 this part really matters. Because job searching 00:07:47.209 --> 00:07:50.870 isn't just about the skill. It's about stamina. 00:07:51.529 --> 00:07:54.910 So when you've been rejected enough times, you 00:07:54.910 --> 00:07:58.269 start internalizing it. Because you start thinking, 00:07:58.470 --> 00:08:03.209 maybe I'm not just good enough. But here's what 00:08:03.209 --> 00:08:06.689 the AI taught me. Rejection is just a data point. 00:08:06.829 --> 00:08:10.110 It's not feedback on your worth. It's feedback 00:08:10.110 --> 00:08:13.689 on your fit. I started to see my own job search 00:08:13.689 --> 00:08:16.750 like a growth model. If a post doesn't convert, 00:08:16.930 --> 00:08:19.810 it's not personal. It's a signal to tweak the 00:08:19.810 --> 00:08:23.709 system. And suddenly it wasn't about hope anymore. 00:08:23.990 --> 00:08:27.970 It was about iteration. AI gave me the distance 00:08:27.970 --> 00:08:31.509 from the emotional chaos. It gave me back control. 00:08:33.190 --> 00:08:37.409 Now let's see the human behind the code. But 00:08:37.409 --> 00:08:39.450 here's the funny part, right? The more I automated 00:08:39.450 --> 00:08:43.389 the process, the more human I felt. Because automation 00:08:43.389 --> 00:08:47.409 didn't mean any detachment. It meant freedom. 00:08:48.509 --> 00:08:53.210 Now, when my AI handled the grunt work, I had 00:08:53.210 --> 00:08:56.889 time to actually research companies, to connect 00:08:56.889 --> 00:09:01.149 with people, to write thoughtful follow -up messages. 00:09:02.639 --> 00:09:05.120 to reflect on which roles truly aligned with 00:09:05.120 --> 00:09:09.340 who I wanted to be. Now, in a strange way, my 00:09:09.340 --> 00:09:13.340 AI system gave me back the emotional energy that 00:09:13.340 --> 00:09:16.519 job searching had stolen. And somewhere in that 00:09:16.519 --> 00:09:18.940 process, I stopped chasing roles and started 00:09:18.940 --> 00:09:22.379 choosing them. And that's when I started seeing 00:09:22.379 --> 00:09:25.639 results. Like interviews came in, conversations 00:09:25.639 --> 00:09:29.440 became easier, because I wasn't coming from desperation. 00:09:30.860 --> 00:09:34.840 I was coming from clarity. If there's one takeaway 00:09:34.840 --> 00:09:40.200 from this episode, it's this. Automation should 00:09:40.200 --> 00:09:42.840 never replace your purpose. It should remove 00:09:42.840 --> 00:09:47.379 friction so that you can focus on it. My AI agents 00:09:47.379 --> 00:09:49.700 didn't get me the job. They gave me back the 00:09:49.700 --> 00:09:53.220 time and confidence to show up better. And they 00:09:53.220 --> 00:09:56.279 reminded me that efficiency isn't about doing 00:09:56.279 --> 00:10:01.600 more. It's about doing what matters. And if you're 00:10:01.600 --> 00:10:04.259 listening to this right now, in the middle of 00:10:04.259 --> 00:10:08.240 your own job search spiral, I see you. You're 00:10:08.240 --> 00:10:11.860 not lazy. You're not even lost. You're just human 00:10:11.860 --> 00:10:14.940 in a system that was never designed to make this 00:10:14.940 --> 00:10:18.659 easy. Maybe it's time to build your own system. 00:10:19.299 --> 00:10:23.759 Not a perfect one. Not a fancy one. Just one 00:10:23.759 --> 00:10:26.860 that helps you show up. One that helps you stay 00:10:26.860 --> 00:10:30.899 consistent, curious and calm. And if that system 00:10:30.899 --> 00:10:34.580 includes AI, great. Because AI isn't the future 00:10:34.580 --> 00:10:39.240 of job hunting, it's agency. And AI just helps 00:10:39.240 --> 00:10:41.600 you reclaim it. The new kind of teamwork that 00:10:41.600 --> 00:10:44.919 I want to talk about here is, this is the story 00:10:44.919 --> 00:10:50.179 really. How I used three AI agents to fix my 00:10:50.179 --> 00:10:54.480 job hunt. Not because I wanted shortcuts, but 00:10:54.480 --> 00:10:59.340 because I needed structure. Like I said, AI didn't 00:10:59.340 --> 00:11:01.799 find me the job. It helped me find myself again 00:11:01.799 --> 00:11:04.600 in the process. And if you want to try building 00:11:04.600 --> 00:11:09.279 your own version, I'll drop my own in the show 00:11:09.279 --> 00:11:13.139 notes. And how, you know, it'll be like a job 00:11:13.139 --> 00:11:16.460 hunt companion for you using AI. And it's the 00:11:16.460 --> 00:11:20.779 same tool I used. Be free and open for anyone 00:11:20.779 --> 00:11:23.919 who wants to make job searching a little less 00:11:23.919 --> 00:11:28.230 soul crushing. I'll also be attaching a blog 00:11:28.230 --> 00:11:33.250 on this link. So monitor this podcast in the 00:11:33.250 --> 00:11:36.669 next week and you'll be getting a link to a blog 00:11:36.669 --> 00:11:39.830 as well with the full architecture, prompt chains 00:11:39.830 --> 00:11:45.509 and sample outputs, even resume examples. Until 00:11:45.509 --> 00:11:48.190 then, keep learning, keep building and don't 00:11:48.190 --> 00:11:52.850 let rejection define your direction. Just to 00:11:52.850 --> 00:11:54.889 give some more clarity on when the blog will 00:11:54.889 --> 00:12:00.289 be out. The plan is to release this episode on 00:12:00.289 --> 00:12:04.149 November 4th. So it will be roughly a week from 00:12:04.149 --> 00:12:08.450 November 4th. And if you're listening this, maybe 00:12:08.450 --> 00:12:11.769 in 2026, just know that this is now about for 00:12:11.769 --> 00:12:15.149 2025 that I will be releasing this episode. So 00:12:15.149 --> 00:12:17.370 a week from November 4th is when you should expect 00:12:17.370 --> 00:12:22.049 the blog post out. I'm Mukundan and this is Data 00:12:22.049 --> 00:12:24.870 and AI with Mukundan. See you next week.