AI Platform Engineer

About the role

This role is for an engineer to build the core execution engine for Supabase's internal AI systems. You will create the platform that powers AI agents, ensuring reliability, safety, and measurable impact across the company.

Key facts

What you'll do

• Develop the production-ready agent platform, including an event-driven queue, a flexible model runtime, persistent state for interrupted tasks, human review checkpoints, and comprehensive logging for all operations.
• Establish and manage the evaluation framework, implementing automated checks and balances to ensure agent trustworthiness before deployment.
• Construct and integrate a variety of AI agents for tasks such as reporting, drafting content, triaging issues, and answering questions, as well as agents that monitor and enhance the platform itself.
• Embed governance rules directly into the system's code, enforcing risk tiers, credential restrictions, and operational boundaries to prevent unauthorized actions.
• Architect the system's communication protocols with people, focusing on efficient, non-intrusive interactions.
• Maintain and improve the platform's tooling, including compilers, validators, and distribution mechanisms.
• Create a system to measure the platform's operational performance and value derived from agent activities.

Requirements

• Proven experience shipping production LLM agent systems that were relied upon by others, with a history of operational use and incident resolution.
• Expertise in designing and implementing evaluation strategies, including golden sets, behavioral assertions, and CI gates, rather than ad-hoc checks.
• Significant experience with deep API integrations into core work systems and authoring MCP servers, understanding their specific failure points.
• End-to-end ownership of infrastructure on GCP using Python, including provisioning, deployment, monitoring, and rollback with infrastructure as code, and a cloud warehouse.
• A strong sense of how software should interact with humans, treating each communication as a valuable resource.

Nice to have

• Public contributions to the AI agent space, such as open-source frameworks, MCP servers, or widely cited writings on agent reliability.
• Experience with LLM observability, cost tracking, and building reporting pipelines from raw logs.
• A track record of building internal platforms adopted by non-engineers, with insights into user-driven improvements.

Skills & tools

• Python
• GCP
• Cloud Warehouse
• Infrastructure as Code

Practical notes

• Fully Remote hiring globally.
• ESOP (equity ownership) provided.
• Tech allowance for work environment setup.
• Comprehensive health benefits (100% for employees, 80% for dependents).
• Annual company off-sites.
• Flexible, asynchronous work schedule.
• Annual education allowance for professional development.