I started in Spring Boot and React at a small agency in Nepal, and spent the last two years going deeper down the stack: queues, clusters, and the unglamorous parts of making an AI product work for real users.
The pattern in my work is taking something that works in a demo and making it survive production. A model-routing layer because the LLM bill was growing faster than usage. Celery workers across GKE and Modal because rendering queued up behind one box. A ClickHouse schema because Postgres couldn't answer questions over billions of rows.
I've been a co-founder twice, so I'm comfortable owning a problem end to end, talking to the people who use what I build, and shipping without waiting for a spec. I'm now looking for a team where I can do that on a product that has already found its users.