Dade Sheng

AI Solutions Architect

I design production agent systems — multi-provider LLM gateways, MCP and A2A surfaces, and scheduled agent fleets — and ship React/TypeScript products on AWS serverless. A Toptal developer since 2020 with years at Amazon, I help teams go from chatbot demos to reliable AI platforms: model routing, tool design, and cloud delivery. I work daily with Claude Code, Cursor, and specialist agents.

AI solutions architect

What I Do

Production Agent Systems

Production Agent Systems

  • ⚡ Design production agent systems: tool surfaces, knowledge access, and scheduled agent fleets

  • ⚡ Use MCP as a control plane so agents can reach tools and knowledge without one-off glue

  • ⚡ Shape A2A-style agent collaboration and Google ADK-style platforms at an architecture level

  • ⚡ Work daily with Claude Code, Cursor, and specialist coding-agent fleets on cloud workstations

LLM Gateways & Model Routing

LLM Gateways & Model Routing

  • ⚡ Design multi-provider LLM gateways: routing, virtual keys, and spend budgets

  • ⚡ Help teams move from chatbot demos to reliable AI platforms — model routing, tool design, and failover

  • ⚡ Keep provider choice a platform decision, not a rewrite, across Claude, GPT, Gemini, and others

  • ⚡ Ship the gateway next to the product so routing, auth, and cloud delivery stay one system

Cloud Infra & AWS Serverless

Cloud Infra & AWS Serverless

  • ⚡ Years at Amazon and AWS shipping global cloud products on serverless infrastructure

  • ⚡ Host and run production sites on AWS; serverless with API Gateway, Lambda, and DynamoDB

  • ⚡ Auth and identity with Amazon Cognito, AWS Amplify, and AWS AppSync

  • ⚡ Data pipelines on Spark/EMR and migrations off large legacy databases — the delivery layer under AI platforms

Full-Stack Engineering

Full-Stack Engineering

  • ⚡ Ship React and TypeScript products end to end, with Node backends and Postgres

  • ⚡ A Toptal developer since 2020 — product engineering from healthcare tools to cloud consoles

  • ⚡ Containerize and deploy with Docker; comfortable across web, APIs, and mobile surfaces

  • ⚡ Care about the interface people use, not only the agent behind it