Description
Summary We've been running a profitable enterprise SaaS for years. It works great, but it has zero AI in it. That changes now. We want to plug Claude/GPT-4 into our existing product: AI assistant, smart search over our own data (RAG), auto generation, streaming chat UI. The stack is Next.js on the front, Node and Python on the back, multi-tenant, real users, real scale. What matters to us: You've done this before on a live production system. Not a weekend demo, not a tutorial clone. You've dealt with the unglamorous parts: keeping tenant data isolated, controlling token costs, handling rate limits, making sure the AI fails gracefully instead of taking a feature down. You should be comfortable with: ● Next.js / React / TypeScript, Node.js, Python ● Claude API or OpenAI API (ideally both, we want to stay provider flexible) ● RAG, embeddings, vector databases ● Streaming responses, function calling, prompt design ● Working inside someone else's codebase without breaking things It would be better if you know LangChain, Vercel AI SDK, Bedrock or Vertex AI, or have shipped B2B SaaS before. How we'll work: You'll get Slack + GitHub access, work on a feature branch, and I’ll review every PR. How to apply: Skip the generic pitch. Start your proposal with the word "Mango" so I know you actually read this, then tell me in a few lines about the hardest LLM integration you've personally shipped and what made it hard. Links to live work beat certificates.