Founding ML Engineer
Infragrid · San Francisco, CA
The role
This role involves training multimodal models on robot data to detect abnormal behavior, understand failures, and evaluate policies. The position sits between research engineering and production machine learning, requiring work on everything from model training to edge deployment. The goal is to turn research into working systems that improve real-world robot operation.
- Location
- San Francisco, CA
What we know that the posting doesn’t say
- Seen 1 day agostill listed on the employer’s careers page
- Posted 32 days agothe first time we saw it
About Infragrid
Infragrid builds Munari, a system that observes what robots see, sense, decide, and do, using that data to help robotics teams improve their systems.
What you would do
- Train multimodal models over video, language, proprioception, and other sensor data
- Build models for anomaly detection, failure classification, and behavioral evaluation
- Integrate and fine-tune vision-language-action models across different robot embodiments
- Develop training methods for limited, noisy, and imbalanced real-world robot data
- Turn failures and recoveries into evals, labeled datasets, and training signals
- Optimize models for deployment on constrained edge hardware
Must have
- Trained modern vision, transformer, diffusion, generative, or multimodal models
- Highly productive in PyTorch, JAX, or comparable ML stack
- Understand model behavior and systems for reliable training and deployment
- Experience with distributed training, large datasets, or inference optimization
- Comfortable moving between research code, data pipelines, and edge deployment
- Excited by messy real-world data without clean benchmarks
- Strong experimental judgment to distinguish genuine improvement from demos
Nice to have
- Experience with imitation learning, reinforcement learning, or diffusion policies
- Experience with VLAs, world models, or sim-to-real
- Experience with active learning, TensorRT, ONNX, Triton, or CUDA
- Experience with Isaac Sim, MuJoCo, or ManiSkill
- Robotics experience or exceptional background in vision or generative modeling
Key skills
- pytorch
- jax
- multimodal models
- transformers
- diffusion models
- generative models
- vision
- anomaly detection
- failure classification
- representation learning
- vision-language-action models
- imitation learning
- reinforcement learning
- distributed training
- inference optimization
- edge deployment
- tensorrt
- onnx
- triton
- cuda
- isaac sim
- mujoco
- maniskill
- sim-to-real
THE ROLE: Own the loop from raw robot experience to a model running on real hardware. Munari observes what robots see, sense, decide, and do. We want to use that data to detect abnormal behavior, understand failures, evaluate policies, identify recurring failure modes, and help robotics teams imp…
Extracted from the employer’s posting. Read it in full on Infragrid’s careers pageApply without filling the form
Approve this role and the application is completed for you, including a résumé tailored to it. You get a confirmation when it lands, and a credit is only spent when a submission is confirmed.
Other roles at Infragrid
- Senior Backend EngineerRemote
- Founding Software EngineerSan Francisco, CA
- AI EngineerRemote
- Founding AI EngineerNew York City, NY
- Forward Deployed Engineer InternSan Francisco, CA
- Software Engineer InternSan Francisco, CA