Home

Running A/B tests on our recommendation system was painful. 6-8 weeks per test, half our users stuck with potentially terrible recommendations, and by the time we got results, user behavior had already changed. I needed something smarter that could adapt in real-time.

Getting SOC2 Type II compliance was honestly more painful than I expected. We thought it would be straightforward - just encrypt everything and call it done. Turns out there’s a lot more to it when you’re dealing with enterprise customers who trust you with their most sensitive data.

I’ll be honest - when I first pitched using diffusion models for material discovery to my materials science colleagues, they looked at me like I was suggesting we use TikTok algorithms to design spacecraft. But after months of tinkering and some surprisingly good results, even the skeptics are paying attention.

I built this wind farm data pipeline as a proof of concept for ingesting sensor data at scale. The setup uses FastAPI, RabbitMQ, and TimescaleDB - all containerized so you can spin it up with one command. Full code is at github.com/daftengineer/data-ingestion Once it is built you can just do docker-compose up --build How It Works The sample...