Forward Deployed Engineer 路 Applied AI & Full-Stack
Bengaluru, India 路 Remote 路 US / EU / GCC hours
I work embedded with clients: I turn what their departments need into working AI systems, and stay on the hook once those systems are live.
6 products 聽路聽 18+ repositories 聽路聽 1,450+ commits 聽路聽 73% of days in client contact
Two problems I would want to be judged on
272 queries down to 7. A read path in a live accounting product. An N+1 that a tree structure multiplied, where the column in question is a native Postgres enum, so the usual bulk shortcut was unavailable. The fix had to come out of the read path itself.
A connection pool that was never reused. Recurring 500s in a live financial core, not reproducible on demand. Persistent connections are normally an optimization; under ASGI every request lands on a fresh thread, so they leak rather than get reused. I measured it instead of assuming, found reuse was exactly zero, and the payments path came back.
Open source
reza gives Claude, Cursor, Codex and Aider one shared project memory, so switching assistants stops meaning re-explaining the codebase.
Building HiringTree at Hiretree Labs.
Open to forward deployed and applied AI engineering roles, remote.
Portfolio 聽路聽 LinkedIn 聽路聽 swebreza@gmail.com