Job description

OUR MISSION Expand human ambition in the physical world. Critical infrastructure is constrained by labor shortages, hazardous working conditions, and operational complexity. Watney builds and deploys autonomous robotic systems that increase the speed and capacity of buildout, starting with data centers. ABOUT THE ROLE At Watney, ML engineers turn a live fleet of robots into world class modals. The fleet produces large volumes of video from real work in the field, and turning that data into a model that performs better on the next deployment is one of the hardest problems at the company. Our Machine Learning team members run training experiments, help curate and label the data behind them, and help evaluate what works on a real fleet. You'll build models that cover the gamut of real world challenges including perception, planning, navigation and manipulation. WHAT YOU’LL DO - Own model quality end-to-end: dataset curation, training, and evaluation. - Work with multimodal sensor data from Cameras, IMUs, LiDar and Odometry to shape the training set as the fleet grows. - Train and evaluate policies, experiment with data, feature and architecture ablations. - Deploy your best results to production and build the systems that guarantee low latency, 24/7 inference on a fleet of robots. - Work with Teleoperations and Fleet Operations to get the data the models need. YOU MAY BE A GOOD FIT IF YOU: - Trained large scale multimodal models on a fleet of GPUs using techniques such as video semantic segmentation, 6d pose estimation, VLAs and BEVs. - Have owned a dataset end-to-end: what to collect, what to label, and what to cut. - Write production training and inference code with Python, PyTorch or JAX, Triton. - Have found and fixed a problem with a training run or a model's real-world performance, and can name the evidence that pointed to it. - Experience working on Cluster management systems including Ray, Slurm or Kubernetes. We’re committed to building a diverse, inclusive team. At Watney Robotics, we welcome people of all backgrounds and identities, and we make hiring decisions based on skills, experience, and potential. If you’re passionate about robotics but don’t meet every requirement, we still encourage you to apply! CURIOUS TO LEARN MORE? Follow us here on X x.com and LinkedIn linkedin.com

Requirements

Must-haves

Own model quality end-to-end: dataset curation, training, and evaluation.
Work with multimodal sensor data from Cameras, IMUs, LiDar and Odometry to shape the training set as the fleet grows.
Train and evaluate policies, experiment with data, feature and architecture ablations.
Deploy your best results to production and build the systems that guarantee low latency, 24/7 inference on a fleet of robots.
Work with Teleoperations and Fleet Operations to get the data the models need.
YOU MAY BE A GOOD FIT IF YOU:
Trained large scale multimodal models on a fleet of GPUs using techniques such as video semantic segmentation, 6d pose estimation, VLAs and BEVs.
Have owned a dataset end-to-end: what to collect, what to label, and what to cut.
Write production training and inference code with Python, PyTorch or JAX, Triton.
Have found and fixed a problem with a training run or a model's real-world performance, and can name the evidence that pointed to it.
Experience working on Cluster management systems including Ray, Slurm or Kubernetes.
CURIOUS TO LEARN MORE?

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

mlmachine learningpythonkubernetes

Platform: ashby
Effort: quick