My wife, Sabrin, is brilliant and incredibly talented at what she does. Like many professionals, she’s always open to new career opportunities, but the process of job hunting is a soul-crushing, full-time job in itself. Every day, she would spend hours searching LinkedIn, tailoring her resume, and writing cover letters, all while managing her current role.
As a Product Manager who builds AI agents, I saw this not just as a problem, but as the ultimate challenge. Could I build an agent that could act as her personal, 24/7 career strategist? An AI that could automate the entire tedious process, from finding the perfect job to preparing the perfect application?
So, I built “Project Taniv-Hawk.”
This is the story of that project. It’s a case study in building a highly advanced, practical AI agent. But more than that, it’s a story about using technology to solve a very human problem. The project was so successful that when a close friend recently lost his job, I realized this tool wasn’t just a gift for my wife—it was a product that could genuinely help people. I’m now in the process of productizing this solution for him.
The Goal: An Autonomous AI Headhunter
The mission was to build a digital twin of a world-class recruiter. It needed to be smart, strategic, and completely autonomous. The agent was designed with three distinct “departments”:
- The “Hunter” (Perception): This department’s job was to scan the internet (specifically LinkedIn, via a professional API) for new, high-quality job listings in her target markets, like the USA.
- The “Strategist” (Reasoning): This is the AI core, powered by Google’s Gemini. It reads my wife’s comprehensive professional profile and the new job description, and then writes a unique, perfectly custom-tailored cover letter for that specific role.
- The “Infiltrator” (Intelligence): A multi-tool system that uses professional APIs like Hunter.io and SerpApi to find the likely email address of the hiring manager for each job.
- The “Executive Assistant” (Action): The final step. The agent takes the tailored cover letter and the contact email and prepares a complete, ready-to-send application draft.
The entire system is designed to run automatically every single day, presenting a fresh list of actionable job applications.
The Challenge: Building a Truly Intelligent Agent
This was the most complex agent I have ever built. The journey was filled with challenges that tested every aspect of my skills.
Problem #1: The Data Problem (Finding the “Right” Jobs)
Simply scraping LinkedIn is a recipe for disaster. It’s a fortress. I quickly learned that the only reliable way to get high-quality, real-time job data was to use a professional, third-party API. This was a crucial strategic pivot from direct scraping to intelligent data consumption.
Problem #2: The “Amnesia” Problem (Avoiding Duplicates)
My first successful version of the agent was a success… for one day. The next day, it would find the exact same jobs and prepare the exact same applications. It had no memory. The fix was to give the agent its own database (using Supabase), which it now reads every morning to learn what jobs it has already processed, ensuring it only ever works on new, fresh opportunities.
Problem #3: The “Lazy AI” Problem (Ensuring Quality)
The AI brain was brilliant, but sometimes it would get lazy and write a generic cover letter. I had to become a “prompt engineer,” writing a new, much more powerful set of instructions that forced the AI to act as a world-class career coach, commanding it to select only the most relevant skills from my wife’s profile for each specific job. This transformed its output from “good” to “exceptional.”
The Breakthrough: A Working “Application Factory”
After weeks of debugging, I ran the final mission. I watched the log as the agent executed its entire workflow flawlessly. It found new jobs, it wrote brilliant, custom cover letters, it found the contact emails, and it saved the final, ready-to-send drafts to its database. The result was a list of perfect, actionable application packages.
TO: [email protected]
SUBJECT: Application: Senior Product Manager...
Dear Deel Hiring Team,
(A perfectly tailored, 3-paragraph cover letter appears here...)
This system now saves my wife hours of tedious work every single week, allowing her to focus on what really matters: preparing for the interviews.
From a Gift to a Product: What’s Next
This project started as a gift, born out of a desire to help my wife. But when my friend reached out for help with his own job search, I realized the true potential of what I had built. The core engine is incredibly powerful and can be adapted for anyone.
I’m now in the process of turning this agent into a real product. The plan is to build a simple interface where any user can upload their own profile and define their own job targets. It’s a reminder that sometimes the most personal projects can have the broadest impact.
This journey taught me that the future of AI isn’t just about big, corporate solutions. It’s about building small, powerful, personalized agents that can act as our advocates, our assistants, and our partners in achieving our goals.