Projects
Things I built outside work to try out ideas.
ModelBench: AI model leaderboard
A public leaderboard of 880+ AI models across 275 benchmarks, with 90% score intervals, prices, and context windows. Rebuilt and redeployed every 12 hours by GitHub Actions; every number links to its source.
- Next.js
- Python
- GitHub Actions
- Data pipeline
- LLM agent
Multi-Agent Conversational Data Analyst
Ask questions about your data in plain language; an LLM turns them into SQL, keeps context across turns, and draws charts.
- NL→SQL
- LLM
- Chart.js
- SQL-injection prevention
- Docker
AI Resume Analyzer
Matches a resume to a job description with Sentence Transformers and cosine similarity, then gives LLM-written feedback. Deployed on Hugging Face Spaces.
- FastAPI
- Sentence Transformers
- Hugging Face Spaces