Naman Kumar

Designed, built, and launched an autonomous deployment diagnosis and error monitoring agent that analyzes runtime logs, failed deployments, code, and configuration across millions of services. Architected evaluation and execution infrastructure for Railway Agents using Temporal workflows and shipped agent billing/usage metering and scalable log fingerprinting via ClickHouse.

Scaled Sourcegraph’s Cody AI coding assistant from prototype to multi-million ARR, serving 250+ enterprises and 15K+ daily developers. Built hybrid multi-repo context retrieval using BM25, vector embeddings, and reciprocal rank fusion, and enabled context and tool execution with MCP; also led Cody Web and Permission Center initiatives and built Stripe-backed monetization plus in-product ROI analyt

Built, launched, and A/B tested 10+ product growth experiments including personalized feeds, content gating, promo codes, and ad placement optimizations. Implemented spam detection that flagged 80% of fake accounts, improved page load times using CDN edge-worker caching, and built event platforms for nomination and voting flows during peak traffic.

Developed a fast, maintainable streaming GraphQL parser adopted in graphql-js and GraphiQL. Improved performance and enabled syntax-aware highlighting.

Built a self-serve chat system to automate customer support, handling about 12K daily tickets during peak sales. Reduced manual workload by 65% and reduced mean response time by 25 minutes.

Built an accounts-payable automation platform that reduced AP operations cost by up to 85%, including an OCR-based document aggregation and auto-indexing pipeline plus an ERP sync agent. Built a cloud kitchen food-ordering platform spanning consumer, cook, and delivery-hub applications for ordering and dispatch coordination.

Led frontend development of FaceCode, a remote interview platform with real-time coding via WebSockets, live video, and code execution. Built enterprise analytics, billing, and team management features including RBAC, invitations, and admin workflows.

Built an online IDE supporting 13 programming languages with real-time collaboration features. Wrote a static site generator that cut code redundancy by 80% and taught Data Structures & Algorithms and React.js to 100+ students.

M.S. in Machine Learning & AI with a thesis on optimizing information retrieval in conversational context via LLM query rewriting. Focused on retrieval improvements for LLM-powered conversational systems.

B.Tech in Computer Science & Engineering with a project on a distributed key-value fault-tolerant in-memory database. Worked on building resilient storage and synchronization mechanisms.