[Job - 32563] AI Lead Software Engineer, Brazil

Role overview

Lead the development of a Python SDK that helps engineering teams build and operate AI agents consistently across frameworks and deployment environments. The role owns the SDK’s public interface and release practices, while shaping orchestration patterns and runtime safeguards for reliable multi-agent systems. It also includes technical leadership, design review, incident follow-up, and mentoring.

Responsibilities

- Own the SDK as a product, including its public API, semantic versioning, deprecation approach, migration guidance, reproducible builds, and release process. - Design adapter and host contracts so agent behavior remains consistent across frameworks such as Strands, Claude Agent SDK, and LangGraph, and across supported deployment targets. - Lead multi-agent patterns such as supervisor/worker and planner/executor/verifier designs, declared-peer delegation, and agent-to-agent communication. - Build runtime protections, including call and hop limits, session budgets, timeouts, per-peer circuit breakers, and actionable structured failures. - Define extension points and convert repeated workarounds into supported SDK capabilities. - Maintain the parity and contract tests that gate SDK and template releases; lead design reviews and root-cause analyses. - Mentor engineers, divide large refactors into coordinated tracks, and deliver against release dates.

Requirements

- Extensive professional software engineering experience, including team leadership and experience building libraries, SDKs, or platforms used by other teams. - Expert Python skills, including asynchronous programming, typing, packaging, dependency management, and pytest at scale. - Production experience building LLM agents with a framework such as Strands, Claude Agent SDK, LangGraph, OpenAI Agents SDK, or CrewAI. - Understanding of agent tool use, context, turn limits, and streaming; experience with MCP and agent-to-agent communication or equivalent. - Experience managing breaking changes in a widely used library through versioning, compatibility measures, or migration tooling. - Strong distributed systems fundamentals, including timeouts, retries, idempotency, circuit breakers, backpressure, and failure handling. - AWS experience, ideally with Bedrock and Bedrock AgentCore; advanced English proficiency and clear technical writing for international collaboration.

Nice to have

- AWS certifications or recent study and project work in agent technology may substitute for some hands-on experience.