Member of Technical Staff, Deeptune Environments
New York City • FullTime
Posted 1mo ago
Remote Work Policy
On-site
Employment Type
FullTime
Categories
Applied AI Engineer
About the job
Mercor is seeking a Member of Technical Staff for their Deeptune Environments team. This role involves building the core systems that power high-fidelity simulation environments for training AI agents through reinforcement learning. You will develop the APIs and tool interfaces agents interact with, the grading layer for evaluating agent success, pipelines for transforming human demonstrations into training environments, and the orchestration and sandboxing infrastructure for large-scale operation. You will collaborate closely with RL researchers, translating their ideas into functional, scalable systems, and will be responsible for the end-to-end development of these critical components.
Responsibilities
- Build systems for end-to-end environment creation, including simulated applications, agent-facing tool surfaces, task definitions, and scoring verifiers.
- Design and operate backend infrastructure for large-scale environment execution, encompassing containers, orchestration, queues, and observability.
- Transform raw human data into clean, reproducible training environments.
- Prioritize reliability and speed alongside correctness in system development.
- Manage the interface with AI labs and researchers, translating research goals into implemented systems.
- Contribute to the development of post-training, evaluation, and reward modeling systems to support RL research.
Requirements
- 2+ years of full-stack or backend engineering experience.
- At least 1 year of experience at a startup, ideally as a founding or early engineer.
- Strong generalist with systems depth, fluent in Python and at least one other language.
- Familiarity with applying agents and ability to adapt to diverse problem requirements.
- Comfortable with ML/LLM concepts, including post-training, evals, and reward modeling, to effectively partner with researchers.
- Ability to thrive in ambiguity, scope work independently, make pragmatic decisions, and ship without detailed specifications.
- Capacity to elevate the skills of other engineers and drive execution while remaining hands-on with code.