some things i have worked real hard on.
python, fastapi, redis, postgresql, sqlalchemy, aws ec2
- built a distributed job queue with priority scheduling, delayed execution and at-least-once delivery - 150 jobs/sec across 4 workers - atomic job claiming with redis lua scripts, so zero duplicate runs across 10,000 test jobs - 60s leases + 20s heartbeats, and a reaper that recovers crashed workers' jobs in under 65s - exponential backoff retries, dead-letter queue, and idempotent submission via postgres unique constraints
react, typescript, fastapi, postgresql, sqlalchemy, gemini ai, supabase, tailwind css, docker
- react, typescript, fastapi, postgresql, gemini ai, supabase, docker - ai knowledge manager for 5 content types - notes, links, pdfs, images, videos - tested with 10+ real users - gemini + vector embeddings handle auto-tagging, semantic search and ai summaries - 30+ rest endpoints with google oauth and jwt, file storage on supabase
next.js, typescript, django rest framework, gemini, groq, docker, nginx, aws ec2
- ai vehicle diagnosis from text, photos, audio or video, with repair reports, severity ratings and cost estimates in inr - hybrid rules engine + decision trees for 6 critical systems resolve 40% of queries without an llm call - cheaper, and the app survives api outages - multi-model fallback chain across gemini and groq - deployed on aws ec2 behind a cloudflare tunnel, frontend on vercel