This role focuses on improving machine learning models using data generated by a live fleet of robots. As a Machine Learning Engineer at Watney, you will conduct training experiments, refine real-world datasets, and assess performance for the next deployment in an end-to-end workflow.

Role Overview

At Watney, ML engineers transform a live fleet of robots into progressively better models. The fleet generates significant volumes of video from field operations, and converting that material into a model that performs better on the next deployment is a core challenge for the company. The modeling effort covers multiple task types, including what the robot observes, the decisions it makes, and the actions it takes.

Your work will support a staged approach that moves from imitation learning toward broader generalization. This includes running training experiments, improving the curation and labeling of the data used for those experiments, and evaluating which approaches perform effectively on the real fleet.

Responsibilities

Run and evaluate training experiments as models scale.
Curate and clean the data used to train experiments.
Support the evolution of labeling processes and data structuring for new datasets.
Track performance metrics that distinguish models that help from those that do not.
Collaborate with Teleoperations to understand data at its source and improve it.

Requirements

Have trained models on real-world data pulled from actual operation.
Have worked with imitation learning, reinforcement learning, or a comparable control method.
Write production ML code using Python and PyTorch (or similar tools).
Be comfortable cleaning and curating messy, real-world data.

Technologies

Python
PyTorch

Location

San Francisco, CA (onsite)

Mission

Expand human ambition in the physical world.

Additional Information

You will help decide the path forward by pairing experimentation with dataset improvements and real-fleet evaluation.

Watney may share updates via X and LinkedIn.

Machine Learning Engineer

Job Description