Senior / Principal Engineer — AI-Powered B2B Go-To-Market Platform (Startup)
~4 years, most recent role
An early-stage startup that pivoted from a low-code backend platform to an AI-powered go-to-market (GTM) platform — software that discovers, qualifies, and engages high-intent B2B leads from real-time behavioral signals using agentic AI. I owned backend, platform/infrastructure, and the AI engine, and drove the platform's evolution from a multi-service MVP to a unified, AI-first system.
- Reliability & scale at the data layer: Designed a dual-database architecture — a managed relational database for transactional multi-tenant data, and a columnar analytics database (OLAP) over hundreds of millions of deduplicated contacts and tens of millions of companies — with async processing on a Postgres-backed job queue for scheduled signal runs, autopilot cycles, and batch enrichment. Built supporting microservices (e.g. an internal IP-geolocation service: Node/TypeScript, offline geo database, memory-mapped reads, timing-safe bearer auth).
- Agentic AI engine: Built the engine that routes work across multiple LLM tiers (a large reasoning model, a mid-tier writing model, and a fast classification model) and cost-efficient open models, using a modern AI SDK with prompt caching and extended thinking to control cost, latency, and quality. Integrated tool-calling servers (web search, enrichment, professional-network lookups, browser automation, email) with strict tool allowlisting to prevent model drift.
Tech: TypeScript, Node.js, Express 5, React 19, PostgreSQL (managed), ClickHouse, Postgres-backed job queue, modern AI SDK, Anthropic Claude, OpenAI, Google Gemini, tool-calling protocols, Python/FastAPI, Docker, modern cloud hosting.