Work with me

Hire

Available from August 2026 for contract and fractional engagements. Based in Gurugram, working with teams in India, the US and Europe.

I work on one problem: GenAI systems that demo well and fail in production.

Nine years in software, the last three shipping GenAI for 150+ enterprise tenants under SOC 2. I have run the RAG assistant that was confidently wrong 35% of the time and brought it to 8%. I have shipped agents with human approval gates and failure-mode guardrails specified before implementation. I have done the multi-tenant isolation work that enterprise procurement asks about.

I also ship my own products. EMpath, an AI-native 1:1 and growth-tracking tool for engineering managers, went from idea to live product solo and runs at useempath.app. When I quote two weeks, I mean two weeks.

Four fixed-scope engagements

RAG Evaluation Audit

1 to 2 weeks

₹1.5L to ₹3L · $2,500 to $4,500

For: Teams whose retrieval-backed answers are wrong and who do not know why.

You get: A working eval harness on your own data, a retrieval failure taxonomy, measured baselines for hit rate and faithfulness, and a ranked fix list with expected impact per fix.

LLM Cost Reduction Audit

1 to 2 weeks

₹1.5L to ₹3L · $2,500 to $5,000

For: Teams whose inference bill is growing faster than usage.

You get: Full token accounting by workflow, a model routing and cascade design, a caching strategy, prompt compression opportunities, and a Bedrock provisioned versus on-demand analysis. Findings quantified as monthly savings.

Agent Architecture Review

1 to 2 weeks

₹2L to ₹4L · $3,500 to $6,000

For: Teams whose agent works in demos and behaves unpredictably with real users.

You get: A topology review, tool-call reliability analysis, an enumerated failure-mode map with recovery paths, an observability and tracing gap assessment, and an MCP integration path if relevant.

Multi-Tenant AI Compliance Review

1 to 2 weeks

₹2L to ₹4L · $3,500 to $6,000

For: Teams selling AI into enterprise procurement and getting blocked on security review.

You get: Tenant isolation assessment across vector stores and retrieval boundaries, PII exposure review across prompts, logs and traces, audit trail design, and a remediation plan mapped to SOC 2 controls.

How it works

  1. 01A 30-minute call. Free. I tell you which of the four you actually need, or that you need none of them.
  2. 02A one-page scope with a fixed price and a delivery date.
  3. 0350% to start, 50% on delivery.
  4. 04A written report and a walkthrough call.

You can act on the report without me. If you want me to build the fix afterwards, that is a separate engagement.

Also available for

Fractional AI architecture, 8 to 12 hours a week, on a monthly retainer.

Subcontracting to agencies and consultancies that need senior AI architecture capacity on an existing engagement.

Full-time: AI Solutions Architect, Forward Deployed Engineer, Principal AI Engineer. India or remote.

Start here

AboutEMpathWritingProductionConnect