Inspiration

GitHub is often treated like a résumé, but no one really tells developers how it actually looks to others. I was inspired by the idea that honest feedback—when delivered with humor—can be far more memorable and useful than generic advice. I wanted to build something that was fun, slightly brutal, but ultimately helpful: a mirror that reflects your GitHub profile the way recruiters, senior engineers, and the internet might see it.

What I Learned

I learned how to balance humor with usefulness—making sure the roasts were grounded in real data and never purely arbitrary. On the technical side, I gained hands-on experience working with the GitHub API, prompt engineering for LLMs, and generating visual assets with AI image tools. I also learned how important pacing, UX, and storytelling are when demoing a project in a hackathon setting.

How I Built It

This was made completely using vibe coding with Gemini AI Studio. GitRoast pulls public GitHub data such as repositories, commit history, languages, and README quality using the GitHub API. I process this data with custom heuristics to generate scores, archetypes, and insights. Gemini is then used to turn those insights into a structured roast and actionable advice. Finally, I generate a shareable “roast card” image using AI image generation, wrapping everything in a clean, dark-mode web interface optimized for quick demos and virality.

Share this project:

Updates