Liam Liu
Overview

Applied AI · Agent orchestration & ops

AgentOps Studio: Agentic Workflow Platform

A tool-augmented workflow platform that lets LLM assistants plan, call tools, and stream results reliably.

20+ concurrent sessions
System architecture01 — 05
01MCP Tool Layer
02Orchestration Engine
03Async Workers
04WebSocket Streaming
05Monitoring & Logs

My scope

The boundary I owned.

I designed the MCP tool layer, the agent orchestration engine, and the asynchronous execution and monitoring backend.

  • TypeScript
  • Next.js
  • FastAPI
  • MCP
  • Redis
  • Docker

Engineering evidence

  1. Built a tool-augmented AI workflow platform on MCP, giving LLM assistants access to 5+ external tools — files, calendars, notes, and task data — through structured tool calls.
  2. Designed an orchestration layer with task planning, tool routing, retry handling, and session context to support reliable 3+ step workflows.
  3. Implemented FastAPI and Redis background workers with WebSocket streaming for asynchronous execution and 20+ concurrent sessions, plus monitoring that cut failure investigation time by 30%.
OutcomeReliable multi-step agent workflows with full execution observability.