Set the technical direction for Compass and our LLM-powered features. Architect multi-agent systems, own the major design decisions, and balance rapid experimentation with stable, scalable long-term architecture.
Lead the design and evolution of the MCP servers, A2A integrations, and APIs that connect internal and external agents to the Basis media buying toolkit.
Own agent orchestration, evaluation, and observability. Define how we measure agent quality and ensure agents reason, plan, and execute reliably in production. Own our software end to end, including production support for Compass.
Serve as the team’s primary technical point of contact for Product, QA, Design, Operations, other Engineering teams, and relevant stakeholders. Represent the team in cross-team planning, architecture discussions, and RFCs.
Drive large-scale technical projects that span multiple teams and systems, from ideation through production. Own team deliverables, decisions, quality, process, and technical direction. Develop a deep understanding of our business domain and stakeholder needs, translate business goals into realistic technical plans, and proactively identify risks and blockers.
Own the team’s technical quality bar through design and code reviews. Onboard new engineers, mentor senior and mid-level engineers, and set a positive example on how to collaborate in pull requests, demos, and working relationships.
Lead the adoption of AI-assisted and agentic development across Basis Engineering by establishing best practices and building the examples and infrastructure other teams adopt.
- Embody Basis’ core principles while fostering a positive culture within the team. Actively promote Basis in the community.
- WHAT YOU BRING TO THE TABLE
- Minimum of 8 years of professional software engineering experience, including recent experience shipping production LLM-powered applications or agent systems.
A track record of technical leadership, including owning the technical direction of a team or major system and influencing architecture beyond your immediate area of ownership.
Expert-level proficiency in Python, with a history of architecting production APIs and backend services and integrating with protocols such as MCP and A2A.
Deep experience building and operating production-grade agentic systems with frameworks such as Strands, LangChain, or the Claude Agent SDK, including asynchronous and event-driven patterns, prompt engineering, structured outputs, and techniques for improving agent reliability. Sound judgment on when to optimize for speed and when to invest in quality.
Experience with evaluation and observability tools such as Langfuse or LangSmith, including designing evaluation strategies from the ground up.
Fluency with AI coding tools (Claude Code, Cursor, Codex, etc.) as a core part of the development workflow, paired with strong judgment and full ownership of AI-generated output.
Proven ability to mentor engineers and to explain complex agentic systems clearly to engineers, Product and Operations partners, and non-technical stakeholders.
- Thoughtful approach to collaboration, design, and decision-making that prioritizes equity, access, and continuous learning, and a commitment to creating inclusive, respectful environments where all voices are valued and supported.
- BONUS POINTS
- Bachelor’s or advanced degree in Computer Science, Engineering, or a related field
Prior experience in ad tech, media planning systems, or programmatic advertising platforms.
Hands-on experience with AWS Bedrock, including Bedrock AgentCore (Runtime, Gateway, Identity).
Experience designing MCP tool sets or skills libraries adopted by multiple teams or external partners.
Experience with authentication and authorization for non-human actors, such as OAuth 2.0 flows for agents and token vaulting for third-party integrations.
- Experience with RAG systems, vector databases, or knowledge retrieval approaches.
- OUR TECH STACK
- Languages: Python
Frameworks: LangChain, LangGraph, FastMCP, FastAPI
Protocols & Integrations: MCP, A2A, AG-UI, REST APIs
Evaluation & Observability: Langfuse, custom evaluation frameworks
Data: Snowflake, S3
Infrastructure: AWS, AWS Bedrock (including Bedrock AgentCore)
