Open to forward-deployed and applied AI engineering roles
Contract or full-time. I work embedded with teams on production LLM systems: scoping fuzzy problems, building the evals, shipping the fix, and staying accountable for the numbers after launch.
How I work
Scope the fuzzy problem with the people who use it.
Discovery with users and stakeholders first, so the work targets the outcome that matters, not the first symptom.
Build the evals before changing anything.
A golden eval set from production traces, so every later change can be shown to hold quality.
Ship the fix in the team's codebase.
Through their review process and CI, gated by evals, with a rollback plan for each change.
Stay accountable for the numbers after launch.
Cost per successful outcome, eval pass rate and latency, measured against the baseline.
Stack
- LLMs: Amazon Bedrock, and direct APIs such as OpenAI and Anthropic
- Evals: golden eval sets, eval gates in CI, LLM judges calibrated against human labels
- RAG, agents, model routing and caching
- AWS, backend systems, CI/CD and workflow automation
- LLM observability and cost: cost per successful outcome, showback, FinOps
Systems I've built
Short write-ups (problem, architecture, evals and outcome) are on the Work page. See the systems
Time zone
- India (IST), working remotely
- Overlap: early morning US East, working hours EU, UAE and Singapore, afternoon ANZ
CV
Email me for a CV. A line or two on the role and the team helps me reply with what's relevant.
[email protected]Prefer a call? Choose "Contract or full-time AI engineering role" in the booking form: Get a free cost check