Senior AI/Backend Engineer — Skills Platform Architecture (MCP)

About Tanren: We build custom Claude-based AI skills for professional service clients. We want to turn our one-off, per-client builds into a shared skills library many small businesses can use — one core, a thin custom layer per business. Client data (names, balances, personal detail) stays fully on their side; what we see and build on is the pattern of decisions and actions taken, feeding a context layer that compounds across clients over time. What's decided: Built on Claude Skills + MCP. Some production skills already exist. Raw client data never leaves the client's own environment — only decisions/actions taken feed back to us. What you'd actually decide: Build our own MCP-fronted library vs. extend a vendor vs. something else How to structure "one core + per-business layer" without forking codebases per client How to design the context layer itself — what decision/action data we capture, how it compounds into an improvement across clients, and exactly where the line sits against raw client data Who we need: Senior judgment, not just execution — someone who lays out real trade-offs and makes a recommendation, comfortable with an underspecified problem. Hands-on: Python and/or TypeScript (the two languages with official MCP SDKs), real experience building MCP servers or agent integrations. Strong plus: multi-tenant SaaS architecture, privacy-sensitive work How we'd start: A short, paid scoping consultation first — then a longer engagement (our guess is a few months, open to your view) if it's a fit on both sides. In your proposal: A time you chose between competing technical approaches and why. Any MCP servers, agent integrations, or multi-tenant systems you've built. Questions this raises for you.