# Research Engineer, Simulation & Reinforcement Learning · Good AI Labs

> Build simulation environments and train policies that improve through interaction, with a clear path to physical robots.

URL: https://www.goodailabs.com/careers/research-engineer-embodied-ai/
Date: 2026-10-13
Last updated: 2026-10-13
Work arrangement: Remote-first


Related work: [Research at Good AI Labs](https://www.goodailabs.com/research/)

## Applying

Email your application to research@goodailabs.com, with the role title in the subject line. Include your location and availability. Please share only material you have permission to disclose.

Email: research@goodailabs.com. Include the role title in the subject line.


Good AI Labs is developing foundational models for robots that learn through experience. We are hiring a Research Engineer to build the simulation and reinforcement-learning systems that make this research possible. You will develop environments, train policies and investigate whether improvements survive changes in the task and the physical system.

Simulation is part of the learning process, not just a place to run a demonstration. You will work across physics, task design, rollout collection and model training. The role calls for someone who can debug both a simulator and a learning algorithm, and recognise when a policy has exploited a flaw in the environment rather than learned the intended behaviour.

## Responsibilities

- Build simulation environments for robot interaction, including observations, actions, contacts, reset logic and task termination.
- Implement and train RL policies; develop rewards, curricula and evaluation conditions that capture the intended task.
- Connect simulated experience to foundational-model training and adaptation, working with researchers on learning objectives and data collection.
- Build parallel rollout, replay and evaluation systems; profile the balance between simulation throughput, training compute and data transfer.
- Investigate sim-to-real transfer through appropriate variation in dynamics, sensing and task conditions, then compare against physical observations.
- Diagnose reward exploitation, unstable training and simulator artefacts; maintain reproducible baselines, held-out tasks and failure analysis.

## Qualifications

- BS, MS or PhD in computer science, robotics, engineering or a related field, or demonstrated technical work of equivalent depth.
- Hands-on experience implementing and training reinforcement-learning policies, with responsibility for the environment, learning loop and evaluation.
- Strong Python and PyTorch skills, with the ability to implement methods from papers and inspect the training code beneath a framework.
- Practical understanding of policy optimisation, exploration, reward design and the effect of observation and action spaces on learning.
- Experience building or modifying physics-based environments and debugging issues involving dynamics, contacts, coordinate frames or timing.
- A record of substantial engineering work that other people can run, inspect and extend, supported by clear experiments and sound software design.

## Preferred experience {#experience-that-would-help}

- Robot manipulation, locomotion, imitation learning, offline RL or adaptation of pretrained models through interaction.
- Physics engines such as MuJoCo or Isaac, GPU simulation, parallel environment execution or distributed rollout collection.
- Domain randomisation, system identification, hardware-in-the-loop testing or deploying learned policies on physical robots.

## Application materials {#what-to-include-in-your-application}

Include your CV and a simulation or RL project you built. Describe the environment, training method, compute setup and evaluation. Explain a failure you diagnosed and what changed after your intervention. Code, training curves, rollout videos and technical reports are useful; clearly identify your own contribution.

Contact: research@goodailabs.com