Pankaj G.

Senior Full-Stack Engineer & LLM Expert | CTO & AI Product Builder I build production-ready LLM-powered applications and have been shipping agentic systems since mid-2022 — before most of today's tooling existed. As a former CTO and two-time co-founder (Mavex AI, acquired), I've architected and shipped multiple AI products end-to-end, from concept to scale. Core expertise: - LLM systems architecture, fine-tuning, RLHF pipelines, and eval harnesses - Multi-agent and autonomous systems — delegation, routing, plan-and-execute loops - Full-stack development with deep AI integrations (Gmail, Google Calendar, Notion, Asana, Jira, Slack, WhatsApp) - Complex backend architectures, API integrations, and async/long-running task systems - Cloud infrastructure: AWS, GCP, Terraform, Docker, CI/CD - MCP servers, codebase indexing, and developer tooling for AI agents Recent work: → First engineer at DeepLaw — built a full contract-analysis pipeline (parallel OCR + extraction agents, legal-entity disambiguation, calendar auto-scheduling) end-to-end in under 3 months, with end-to-end analysis under 50s on 100-page documents → Built trie — a symbol-level codebase index exposed over an MCP server that cuts AI agent tool calls by 3x → Built FlowPilot — a browser-automation agent that replays learned actions with zero LLM calls (10x faster on warm runs) → Architected a multi-agent productivity platform at Mavex AI with dynamically rendered UIs and an AI scheduling assistant CC'd directly into email threads → Built eval harnesses for production AI, reducing extraction noise from 112 to ~18 through iterative prompt engineering I treat models as components with distinct cost, latency, and reliability profiles — not interchangeable black boxes. My focus is scalable, production-ready solutions that solve real business problems. Let's discuss how I can bring your AI project to life.