Minimum qualifications:- Bachelor's degree in Computer Science, Robotics, or equivalent practical experience.
- 2 years of experience with machine learning tools and algorithms, specifically deploying Large Language Models (LLMs)/Vision Language Models (VLMs) and deep learning models.
- Experience in a technical role (software engineering, AI/ML engineering, or solutions architecture).
- Experience with Python, and with modern AI-assisted development tools to accelerate prototyping.
Preferred qualifications:- Experience with ROS/ROS2, or on-device deployment constraints (Jetson, TPU).
- Experience managing large-scale multimodal datasets, time-series telemetry data, or building automated pipelines for hardware-in-the-loop testing.
- Familiarity with the operational realities of modern vision-language-action (VLA) models or behavior cloning policies and their common pitfalls like task overfitting.
- A deep-seated interest in the future of embodied AI and a desire to build the testing bedrock for robotics development.
About the jobAs a Research Engineer, you will manage the practical challenges of benchmarking foundation models for robotics. You will have an understanding of how modern robotics foundation models work and where they currently fall short. Your mission is to design evaluation protocols, tooling, and frameworks that extract meaningful signals from the messiness of physical policy execution. You will build the infrastructure that allows the engineering team to effectively hillclimb and gives leadership a clear, data-driven understanding of technological readiness.
US: $174000 - $253000 (USD) 15% bonus target equity benefits
Learn more about benefits at Google .
Responsibilities - Design, implement, and maintain scalable, robust frameworks to enable large-scale evaluation of robot policies across offline open-loop testing and real-world hardware evaluations.
- Partner with researchers to design the content of various benchmarks in order to maximize evaluation signal and stress-test model capabilities.
- Build diagnostic and visualization tools that allow the team to easily root-cause policy failures and track performance regressions.
- Establish evaluation criteria for model releases and own the stability and benchmarking of models slated for critical demos.
- Innovate on how to make real-world hardware evaluation faster, more reproducible, and less reliant on manual human intervention.
