I build production LLM systems, and measure what they cost and how well they work
AI engineer helping Seed–Series B SaaS teams ship LLM features, evals and the systems around them, with cost measured from day one.
How I think
Measure cost per successful outcome, not cost per request.
A cheap request that fails costs more than an expensive one that works.
No eval, no cost cut.
Every change is checked against an eval before it's recommended.
Architecture before pricing.
Context size, retries and routing set the bill before the price per token does.
Fixes ship in your codebase, not in a slide deck.
A finding counts once it's merged, gated by evals and measured against the baseline.
Background
10+ years building SaaS and AI systems, working directly with the people who use them.
- Production LLM systems: RAG, agents and evals on AWS Bedrock and direct APIs
- Backend and cloud architecture on AWS
- CI/CD and workflow automation, including eval gates
- Customer-facing delivery: discovery, stakeholder work, owning the result after launch
Credentials
FinOps Certified Practitioner
FinOps Foundation
Where I work from
- India (IST)
- Calls Mon–Fri, 9 AM–7 PM IST: early morning US East, working hours EU, UAE and Singapore, afternoon ANZ
Open to forward-deployed and applied AI engineering roles
Contract or full-time, working embedded with teams on production LLM systems. Email me for a CV: [email protected]
Stack, systems and time zone