FeedMeJD

Inspiration

Honestly, I built this because I needed it. During my own job search, I was drowning in tabs and messy spreadsheets. All the job descriptions started to look the same.

But what if, instead of reading JDs one by one, I could treat them like data points? What if my browser could become an intelligent partner, helping me see the patterns and uncover the hidden "skill blueprint" for my career?

So I decided design and build a tool that felt less like a spreadsheet and more like a discovery mission. That's how FeedMeJD was born.

What it does

FeedMeJD is like your personal AI cartographer for your job search. It's a Chrome Extension with a friendly AI pet that "digests" LinkedIn job descriptions for you.

Here's the magic:

• You collect JDs: As you browse, you "feed" interesting jobs to your pet with a single click.
• It finds the patterns: Behind the scenes, it uses on-device AI (Gemini Nano) to extract key skills from each job.
• You get a blueprint: After collecting a few jobs, you unlock your Skill Blueprint—a beautiful, interactive dashboard that visualizes the most in-demand skills for your target roles.

It turns the noise of job hunting into a clear, data-driven signal, showing you exactly what to focus on.

How we built it

I built FeedMeJD with a "fun meets function" philosophy.

• The core experience is built around the AI pet, making the process of collecting data feel delightful and zero-friction.
• The "brain" is the chrome.ai Prompt API, which I used to interact with Gemini Nano. I spent a lot of time on prompt engineering to ensure it could reliably categorize skills into "hard" and "soft".
• The dashboard is where the insights come to life. I used D3.js to create four unique and interactive visualizations: a Word Cloud, a Constellation, a Prism chart, and an experimental Galaxy network graph.

This gives users multiple ways to explore their own professional DNA.

Challenges we ran into

Building on the cutting edge is always an adventure!

• "SPA Hell": Making the pet reliably appear and disappear on a complex Single-Page App like LinkedIn was a major challenge.
• Taming the AI: Getting a consistent, perfectly structured JSON output from an LLM is an art. It took a lot of trial and error with my prompts to make the analysis reliable.
• The D3.js Rabbit Hole: I'm proud of our four visualizations, but each one was a deep dive into the world of data visualization.

Accomplishments that we're proud of

• A truly useful AI tool: It's not just a demo; I built something i'd genuinely use myself. It solves a real pain point in the job search process.
• 100% On-Device & Private.
• Four Beautiful Visualizations: Creating not one, but four distinct and polished ways to see your data was a huge accomplishment. I believe the way you see your data is just as important as the data itself.

What we learned

My biggest lesson was the journey of transforming a simple idea into a polished, valuable product. I learned that the last 20% of the work—the polishing, the fine-tuning, the relentless focus on the user experience—is what makes the difference between a good project and a great one.

While the technical challenges were fun, what I truly enjoyed was thinking like a user. I obsessed over questions like, "Does this feel magical?" and "Does this genuinely make the user's life easier?". That end-to-end process of building something with real value was the most rewarding part of this hackathon.

What's next for FeedMeJD

This is just the beginning of the skill discovery journey haha:

• Resume Matching: Upload user's resume to see how you stack up against your Skill Blueprint.
• Cross-Platform Support: Bring FeedMeJD to Indeed, Glassdoor, and beyond.
• Smart Job Recommendations: Get job suggestions based on your unique skill profile.

Built With

• d3.js
• gemininano
• promptapi
• typescript
• vite