Applied AI Engineer, Codex Core Agent
Role details
Job location
Tech stack
Job description
The Codex Core Agent team builds the kernel of Codex. We own making the agent better, accelerating research, and making those improvements real in production for our users.
That means working across the systems that make Codex actually function as an agent in the real world: the production performance envelope around tokens, latency, reliability, cost, and capacity; the core execution loop and interfaces that turn models into useful behavior; the shared infrastructure that enables other teams to build on Codex; and the feedback loops that turn real-world usage into better models and better agent behavior over time.
About the Role We're looking for applied AI engineers to help bring Codex agents from impressive demos to dependable tools. This role is about improving agent performance on real software engineering tasks and closing the gap between research capability and real-world usefulness.
You'll work closely with research, infrastructure, and product to ensure agents are not just powerful, but useful, steerable, and reliable in practice. The job is not only to improve model behavior in isolation, but to turn those improvements into measurable gains in solve rate, usefulness, and economic value for users.
What You'll Do
- Design and iterate on agent behaviors across real-world coding tasks and long-horizon workflows.
- Work closely with research to develop and run evals to measure agent performance, regressions, failure modes, and edge cases.
- Improve performance through prompting, tool-use strategies, context construction, and model-facing experimentation.
- Analyze failures in production and systematically improve robustness and reliability.
- Build feedback loops and data systems that get better real-task data into evaluation and research.
- Work with product teams to shape user-facing agent experiences and the interfaces the agent depends on.
- Help define what "good" looks like for agents completing complex tasks end-to-end.
Requirements
- Have experience building or shipping machine learning or LLM-powered products.
- Are strong in Python and comfortable with modern ML tooling.
- Have worked on model evaluation, fine-tuning, or prompt design.
- Think in terms of systems and user outcomes, not just model metrics.
- Enjoy debugging messy, real-world failures and turning them into improvements.
- Want to work in the layer that turns research and model potential into systems that actually work for users.
Bonus Points
- Experience with agent frameworks or tool-using LLM systems.
- Research experiencewith code generation models or developer tooling.
- Experience working with large, messy datasets or production logs.
