Research Engineer/ Scientist

About the role

The AI Research team designs studies and analyzes results to advance robotic capabilities. Emphasis is placed on reproducibility, clear reporting, and documentation of methods and results. The role targets candidates who can connect rigorous research to real hardware. You will own the end to end research cycle from hypothesis to deployment on physical systems. You will ensure that every experiment is recorded and that results can be independently verified. You will translate high level ideas into testable predictions on actual robots. You will work closely with engineers to align research objectives with practical constraints.

Key facts

What you'll do

Real hardware performance is tested through extensive validation of deep learning models on physical robots to confirm behavior under real conditions. These tests ensure that practical performance meets expectations and safety standards. Simulation outcomes are verified using MuJoCo, Isaac, or Drake before any real-world testing begins so that models are trusted when interacting with the physical world. Physical-system testing only follows reliable simulation validation. The codebase must handle demanding robotic learning tasks. Complex model architectures, including Vision-Language-Action models, video generation systems, and large-scale transformer architectures, are implemented and tested. Each research outcome can be reviewed and built upon because detailed records of methods and results are maintained. Clear reporting supports reproducibility across experiments. You will prepare detailed documentation that allows others to reproduce your work exactly. You will present findings to both technical and non technical stakeholders using clear and precise communication.

Requirements

A PhD or Master's in stated fields is required for this position to ensure candidates have the necessary research foundation. Candidates must possess over five years of experience deploying deep learning models onto physical robots, prioritizing real hardware deployment over simulation only. Demonstrated work on real robotic systems is expected, with hands on experience in Vision-Language-Action models, video generation, or large-scale transformer architectures required for success in this role. Simulation validation using MuJoCo, Isaac, or Drake is mandatory before any real-world testing occurs. Production-ready code must be written in Python and PyTorch (or JAX) to meet engineering standards. Defensible decisions are ensured because candidates must prepare to defend every methodological choice during evaluation steps. This practice ensures that decisions are well justified and technically sound. You must maintain meticulous records of experiments, parameters, and results to support audits and reviews. You must be comfortable working in a fast paced environment where requirements evolve based on empirical findings.

Practical notes

Typical interview steps

Research interviews usually include a presentation of past work, a technical discussion, and sometimes a research proposal exercise. Candidates may be asked to design a study or critique a method. Depth of understanding is tested more than speed. Interviewers often ask you to present your past work in depth. Being ready to defend every methodological choice is the core preparation.

Good to know

Questions to ask

Useful questions for the interview: what a typical week looks like, how work is assigned, what tools the team uses, and how feedback works. Asking how the role has changed recently and what the team wishes it had known when joining is also reasonable. Questions about the manager's priorities are especially valued.

Career growth

Research careers grow from junior researcher to senior scientist, principal, and lab or research director roles. Some people move into applied research and product work. Publication record or demonstrable impact drives progression, depending on the setting. Research careers reward a strong publication or delivery record. Applied research roles value impact on products as much as novelty.