Forward Deployed Engineer with 4+ years delivering production software into client environments — healthcare, fintech, enterprise IT. I take the ambiguous problem, integrate with whatever the customer already runs, and leave something live behind.
Now doing the same with LLMs in production: agentic tool calling, RAG pipelines, voice AI, code-execution sandboxes. Nine products live right now, built solo.
Forward deployed work isn't a different stack. It's the same engineering, run against a customer's constraints instead of your own.
The stated requirement is rarely the actual one. Ask until the constraint that matters shows up — cost, latency, compliance, or a system nobody wants to touch.
Decide what can be live in two weeks versus two quarters, and say so plainly. A working slice beats a complete plan.
Their SQL Server, their FHIR server, their payment gateway, their VM estate. Meet the system where it is.
Deploy, instrument, alert, hand over. Monitoring and recovery tooling is part of the delivery, not a follow-up ticket.
None of these were mine. Each one came with its own auth, its own schema, its own failure modes, and no option to rewrite it.
Integrated EPIC FHIR R4 APIs into a FastAPI patient platform, with an NLP assistant at 95%+ intent accuracy and AWS Lambda pipelines handling 10K+ monthly events.
Production ETL across API-to-API and database-to-database boundaries, thousands of HR records a day, with scheduling, encryption, and automatic error recovery.
Multiple gateways integrated behind loan-qualification APIs written in Rust and Python, plus subscription billing that lifted revenue 30%.
A FastAPI service driving full VM lifecycle across 50+ vSphere instances, cutting provisioning time 80% while staying audit-compliant.
Automated ticket workflows across engineering teams through REST APIs, removing 60% of the manual effort without changing how anyone worked.
Voice assistant at 90%+ accuracy, re-architected offline-first once cloud cost became the blocker — 60% cheaper, and the data stopped leaving.
Same loop a forward deployed engineer runs — scope it, build it, deploy it, keep it up — except here I was also the customer, the ops team, and the support desk.
Interview-prep platform serving 10,000+ questions at 5K+ RPS with sub-10ms p99 latency. LLM-driven content generation pipeline behind it.
devinterviewmaster.com →System-design learning platform with a Judge0 code-execution sandbox, Razorpay webhook billing, and an LLM voice tutor.
architechiq.com →K-12 EdTech platform with gamified quiz arenas, JWT role-based auth, and a React client in a git-submodule monorepo.
vidyaarena.org →Resume builder with OAuth, a Puppeteer PDF service, and an LLM ATS-scoring engine that returns section-level feedback.
resumebanao.com →Automation service where Playwright workers auto-apply across job boards, with an LLM layer ranking roles and full queue orchestration.
jobs.mehtasystems.dev →Four further services running quietly — content pipelines, internal tooling, and the infrastructure that keeps the rest of this list online.
tanuj.mehtasystems.dev →Two services organisations and one product team — named customers, real deadlines, and environments I didn't get to choose.
Highlighted items are the ones a forward deployed role tends to lean on hardest.
Open to forward deployed, applied AI, and senior backend roles — remote or Gurugram. If you have a customer with a hard integration and a deadline, that's the work I like.