Research Engineer, RL Env New York · San Francisco →

Mecka; New York, Northern (NY, KY); Hybrid; USD 110,000 - 170,000; Full time

Job summary

Mecka AI is seeking a Research Engineer, RL Env to build reinforcement learning environments that enable researchers to train models and assess capabilities. You will translate real tasks into computable problems, implement environments for model training and verify that progress reflects useful behavior.

In this hands-on role, you will write working software, investigate failures, and develop methods with product colleagues and engineers, contributing to reusable environments and evaluation

Qualifications

• Hands-on experience formulating RL problems and building environments.
• Strong debugging and software engineering skills.
• Ability to design controlled experiments and analyze results.

Responsibilities

• Build RL environments with defined tasks, observations, actions.
• Develop rewards and scoring mechanisms.
• Run learning experiments and compare policies.
• Ensure evaluation reliability and report uncertainties.
• Collaborate with researchers and engineers to transfer prototypes.

Skills

RL environment design

Python

Reinforcement learning

Software debugging

Experiment design

Education

PhD or MSc in CS/ML

Tools

OpenAI Gym

PyTorch

NumPy

Job description

About Mecka AI

Mecka AI is building the data infrastructure layer for robotics and embodied AI.
We design and operate global systems for data capture, data labeling, and hardware-enabled workflows used by leading AI labs and robotics companies to train and validate humanoid and embodied AI systems.
We work closely with frontier robotics teams to bridge real-world data, simulation, learning-based systems, and deployed hardware.

The Role

The Research Engineer, RL Env role at Mecka AI involves building reinforcement learning environments that help researchers train models and understand their capabilities. Across Mecka's Labs team, you'll translate real tasks into computational problems, implement environments for model training and test whether measured progress reflects useful behavior.

This is a hands-on research engineering role for someone who knows how to construct reinforcement learning environments. You'll write working software, investigate failures and develop methods with product colleagues, domain experts and engineers.

What you will be doing:

• Build environments: Define tasks, observations, actions and state transitions. Implement reset behavior, termination conditions and measurable outcomes in environments agents can interact with.




• Develop rewards and scoring: Translate task objectives into feedback and evaluation criteria. Test whether agents can exploit scoring rules without completing the intended task.




• Run learning experiments: Implement baseline agents, train and compare policies, and design controlled experiments that isolate the effects of data, methods and environment changes.




• Make evaluation reliable: Separate training and held-out tasks, check for leakage, version experiments and repeat runs. Report uncertainty and performance across conditions alongside aggregate scores.




• Investigate failures: Inspect trajectories and learning behavior to distinguish policy limitations from data, reward or environment problems. Use findings to prioritize the next experiment.




• Build with the team: Work with domain experts to validate task assumptions and with engineers to turn research prototypes into reusable environments, evaluation tools and documented methods.

What you bring:

• RL environment expertise: You have hands‑on experience formulating problems, constructing environments, training agents and critically assessing results.




• Strong programming and software debugging skills; able to build and test research systems that other people can run and extend.




• Sound experimental design and statistical reasoning, including controlled comparisons, evaluation splits, variability and the limits of benchmark results.




• Ability to reason about environment dynamics, reward design and agent behavior, and trace unexpected results to concrete causes.




• Independent research judgment and clear communication; learn unfamiliar domains, work with specialists and explain assumptions, tradeoffs and findings.

Even better if you have:

• Experience building interactive environments, simulators or benchmarks used by other researchers.




• Work on agent evaluation, reward design, imitation learning or learning from real-world data.




• Research artifacts with reproducible experiments, useful baselines and evidence of investigating failures beyond headline scores.