Fullstack AI Engineer.
I build products end-to-end. I've built & maintained production AI applications for businesses: from RAG and indexing pipelines, agents, to workflow automation at scale since 2024. I care about reliability, guardrails, and applicability.
Based in Bhubaneswar, India.
Professional Work
- Dec 2025 — Present
Software Engineer·Full-time
Built the codebase's first true agent harness for a research workflow and a voice dictation input with ASR + fast context-aware output correction (à la Wispr Flow). Fully rewrote the deals layer with a Celery service with batched emails, and the agent-output review surface. Shipped Ask AI for account queries and embedded Oliv inside Salesforce, HubSpot, and NetSuite.
- Nov 2024 — Nov 2025
Founding Engineer·Full-time
Migrated a Next.js app to a monorepo with a Chrome extension. Built a shared tRPC layer for typesafe ops across all clients. Rewrote 45 LLM workflows with the Vercel AI SDK. Grounded copilot answers in user docs via Pinecone RAG. Shipped a Slack bot with overnight cron batch-analysis.
+207,797/-50,345 - May 2024 — Oct 2024
Founding Engineer·Full-time
Sole engineer on the entire product: no agents, no designers. 11 fully-typed listing types, a reverse auction with proxy auto-outbidding and snipe-guard, a winner-cascade flow with variable-rate Stripe commission, and 14 React Email templates with Twilio OTP auth.
+129,843/-65,862 - Dec 2023 — Apr 2024
Fullstack Engineer·Contract
Shipped Stripe credit-based billing with subscriptions and top-ups. Built a test-call generating AI scripts from Q&A pairs via Bland AI. Added a Do Not Call registry with Twilio OTP and audit logging, plus a call analytics dashboard with recording playback.
+111,046/-84,935
Personal Work
All-encompassing AI personal assistant running on a multi agent harness, with workspace-wide RAG across Gmail/GitHub/Linear/Notion; background agents and more.
State-of-the-art production code review agent with Merkle-backed incremental indexing, multi-agent AI harness, OSV vuln triage, issue memory, eval suite, and false-positive tracking.
Writings
Prompt caching and context optimisation for harnesses
Extremely scalable solutions to context bloat and hitting prompt caches.
Multi agent harness architecture
Context pollution, a boss model with a swarm of cheap workers, and how the tool surface got lazy: from loading whole integrations to searching a tool catalog.
Practical AI-assisted programming
Invaluable tips from my workflow for being productive with LLMs without losing your mind. With real examples from my work
Connect
Write me at yashgouravkar@gmail.comor find me on · · · Resume.