All jobs
AI Platform Engineer
About this role
Supabase is seeking an AI Platform Engineer to develop the execution layer for its internal AI systems. The role involves building a central platform that runs agents, including an event-driven queue, a model-agnostic runtime, durable state, a human review gate, rollback, and comprehensive logging. You will also create the evaluation layer with safety criteria and governance, and build various agents across organisational levels while enforcing strict governance, ownership, and human gates. This is a single-engineer role responsible for end-to-end infrastructure, instrumentation, and reporting of the platform’s value. The position targets candidates with experience shipping production AI agent systems and owning infrastructure end to end, preferably on Python and GCP, with a focus on clear governance and observability.
What you would do
- Ship the agent platform to production with an event-triggered queue, headless runtime, durable state, human review gate, atomic rollback and full run logging.
- Own the evaluation layer with behavioral assertions, safety cases, and CI gates to prevent regressions.
- Build and register a portfolio of agents for different roles and a meta layer to observe and improve the platform.
- Enforce governance in code through risk tiers, least-privilege credentials, tool-permission gates, and a comprehensive audit log.
- Design how the system contacts people, including interruption budgets, message bundling, and useful interventions before asking for action.
- Own platform tooling such as the compiler, validator, inventory integrity, and context/capability distribution paths.
- Compute operating measures from production data to produce defensible maturity grades per team.
- Instrument the platform to log and measure its own return, including a monthly ledger of work absorbed by agents.
- Develop infrastructure to support scalable agent expansion and ensure architecture decisions are closed by you as the engineer.
- Provide a clear design and implementation approach for enabling remote teams to adopt and use the platform effectively.
What they are asking for
- Shipped production LLM agent systems with real users and at least one incident you can discuss.
- Experience designing evaluations with golden sets, behavioral assertions, rubrics, and CI gates.
- Experience with deep API work related to system work and MCP servers.
- Owns infrastructure end-to-end in Python on GCP, with infrastructure as code, deployment, monitoring and rollback capabilities.
- Strong understanding of designing safe human contact patterns and governance around AI actions.
Listing details are summarised from the source. Always confirm the role and the employer on the original posting before applying, and never pay to apply for a job.