Responsibilities

Design and own the agent runtime and orchestration layer
Build long-horizon agent workflows: prompt → plan → generate → run/validate → repair → publish
Develop robust evaluation and quality loops including eval harnesses, regression testing, and failure taxonomy
Design model strategies including routing, benchmarking, reliability improvements, and cost/latency optimization
Create debuggable agent systems with tracing, metrics, alerts, and observability

Requirements

Experience building agentic systems involving tool use, orchestration, retry/repair loops, and context management
Comfortable working with modern agent frameworks such as LangChain, LangGraph, Google ADK, Pi-mono, or Vercel AI SDK
Strong engineering skills in Python or TypeScript
Experience shipping and operating production systems
Strong product instincts and ability to translate user experience goals into system design decisions

Built a coding agent, devtool, IDE assistant, or code generation pipeline
Experience building evaluation frameworks (automated scoring, regression suites, release gating)
Experience designing MCP-style tool protocols or robust tool interfaces for LLM systems