Description
Summary Backend engineer (TypeScript/Next.js) to build AI agent SaaS. We're building a web based MVP for youth athletes: an AI chat agent product where each user gets their own isolated agent (per-user memory, data, and context). We need a backend engineer to own the server side end-to-end and work with the frontend engineer to integrate backend and frontend. WHAT YOU'LL BUILD - API routes, agent logic, auth, and hosting (Next.js/TypeScript on Vercel) - Postgres data layer on Neon: user records, agent knowledge base, strict per-user data isolation. - Claude Sonnet integration via Azure AI Foundry, including a tool-calling layer for in-chat widgets. - Safety flow: detect Azure prompt-filter refusals, end the chat, log the triggering message for telemetry, surface a warning card. - Scheduled background jobs aggregating per-user chat logs into reports - Wiring backend routes to the frontend MUST HAVE - AI coding tools as your primary workflow (Cursor, Claude Code, Codex, or similar) - Production experience with TypeScript, Next.js, Vercel, and Postgres (Neon a plus) - Shipped a product with per-user agent/data isolation or multi-agent management - Built an agent with persistent memory and tool calling NICE TO HAVE - Mastra (or similar agent framework), Honcho (or similar agent memory) - Azure AI Foundry - Experience with sensitive data / products for minors (COPPA-adjacent) DETAILS - Timeline: MVP target of 4-6 weeks - You'll work with me and frontend dev directly; scope doc and designs are ready TO APPLY Briefly describe one agent system you shipped (per-user isolation + tool-calling setup). If it's a fit, I'll invite you to a short 15-minute intro call on Upwork before hiring.