Research Scientist, Robot Foundation Models
Mind RoboticsPalo Alto, CA
HirQ verdict
41/100 · Weak. Based on age, reposts, pay, the company's hiring and ghost signals.
- True age
- 298 days
- Reposts
- Posted once
See the evidence behind this score.
Pro shows each part of the score, the reasons, the pay percentile and the listing history.
HirQ's ghost-risk estimate and company labels are automated opinions based on what we observe on employers' public job boards: how long a listing has been open, whether it was re-listed, wording in the description, and how often roles on that board close. They are not statements about an employer's intent, and a high-risk listing can still be a real opening. Figures are as of the date shown. How we measure
At a glance
- Pay
- No pay listed
- Level
- Mid level
- Type
- FullTime
- Team
- Software Engineering
- Work setup
- On-site
Pay vs market
Employer-posted ranges for Science & Research roles in CA in the last 12 months (747 postings).
This job: No pay listed
Mind Robotics hiring
- Open roles
- 35
- New in 30 days
- 4
- Closed in 30 days
- 3
About the role
From Mind Robotics's careers page, checked Nov 20
ABOUT MIND
Mind Robotics is building universally capable robots to transform industrial work, starting on the factory floor. We're vertically integrated:
- We train our own foundation models.
- We build our own robots.
- We collect data directly in the factories where our robots will work, starting with Rivian, our first manufacturing partner, in addition to third-party data.
ABOUT THE TEAM AND THE ROLE
We train robot foundation models that learn how the physical world behaves from large volumes of video and human demonstration, learn actions from teleoperation and on-robot data, and then adapt to specific stations on the line. You'll design and run that training, and you'll watch your models run on our robots in a production factory.
This is a research role with a production bar. Many of our open questions are open across the field: how much web-scale video adds on top of factory demonstrations, how tactile sensing changes what cameras need to see, and how quickly a model can adapt to a new station. You'll help answer them and ship the answers.
We're a flat organization, and everyone contributes directly to getting our robots deployed.
RESPONSIBILITIES
- Pre-train and post-train multimodal robot foundation models (vision, language, proprioception, force and tactile, actions) at large scale.
- Own experiments end to end: data mixture, training run, evaluation, and deployment on real robots.
- Shape data strategy: what to collect at our deployment sites (human demonstration, teleoperation, recovery data), how to combine it with third-party data, how to measure quality, and which mix actually moves performance.
- Build evaluation that predicts real-world success: offline metrics, simulation and on-robot trials tied to station-level task success.
- Work on the robot: debug failure modes on the line and turn model outputs into smooth, reliable actions.
- Help set our research agenda and write up what you learn.
- Build whatever you need to succeed, including your own data and training infrastructure. You'll have strong colleagues to lean on, but you own the problem end to end.
REQUIREMENTS
- You've designed model architectures and trained large models end to end (LLMs, VLMs, video or world models, or robot policies), and can explain what went wrong in your runs as well as what worked.
- Deep understanding of modern ML, training dynamics and optimization at scale, and you keep up as the field moves.
- Experience with distributed systems and data pipelines for large-scale training.
- Hands-on multi-node training in PyTorch or JAX, and comfortable building the data and training tooling you need.
- Strong Python and solid engineering; you write research code others can build on.
- You want to work hands-on with robots, not just with models.
NICE TO HAVE
- Robot learning: imitation learning, VLAs, RL, sim-to-real or dexterous manipulation.
- Models deployed on physical systems under latency and reliability constraints.
- Tactile or force sensing, multi-view video, or world models.
- A PhD, publications or open-source work in robot learning or foundation models.
WHY MIND
- Real deployment. Your model is graded on a production line, with a partner committed to scaling it.
- Data no one else has. Data collected directly in the factories where our robots work, alongside third-party data.
- Compute isn't the bottleneck. Large-scale GPU capacity for a small team, so you won't have to fight for GPUs.
- Early ownership. Meaningful equity and a direct say in architecture and research direction.
Skills mentioned
- Python
- Distributed Systems
- Machine Learning
- LLMs
- PyTorch
- Robotics
Similar jobs
Same role or field, nearby or remote, ranked by HirQ score.
Space Hardware - Postdoctoral Researcher
LLNLLivermore, CA, US
Pulsed Power Laboratory Operations Engineering Associate/Senior Engineering Associate
LLNLLivermore, CA, US
More at Mind Robotics
Systems Engineer
Mind RoboticsPalo Alto
Tech Lead Manager
Mind RoboticsNormal, IL
Keep exploring
More jobs at Mind Robotics
- Tech Lead, TeleoperationPalo Alto
- Head of Robotics Data Acquisition & StrategyPalo Alto
- Systems EngineerPalo Alto
- Tech Lead ManagerNormal, IL
- QA Reviewer, Data AnnotationPalo Alto
- Tools and PLM EngineerPalo Alto
- Electrical Design Engineer, Power ElectronicsPalo Alto
- Field Technician, Data CollectionNormal, IL
Explore similar jobs
open jobsWhat this role pays
pay pointsCompany
open jobsResearch Scientist, Robot Foundation Models
No pay listed · HirQ 41