Software Engineer, ML Tooling and Infrastructure

Position: Staff Software Engineer, ML Tooling and Infrastructure
This is a hands-on software engineering role on a fast-paced applied AI team. Your mission is to build the tooling, pipelines, and infrastructure that bridge the gap between experimental prototypes and production-grade solutions deployed on our robots. You will have high autonomy to tackle a variety of complex engineering challenges, and your work will have a direct and immediate impact on the capabilities of the Atlas robot.
**** What You'll Do:
***** Architect and Refactor:
Take ownership of our Python-based training and inference infrastructure, relentlessly improving its quality, performance, and scalability.
* Build with Quality:
Implement comprehensive testing, champion best practices for code quality, and build automated CI/CD pipelines to ensure reliable deployment and validation.
* Own MLOps:
Design, build, and operate the MLOps infrastructure for our cutting-edge behavior models, focusing on reliability, reproducibility, and speed from training to deployment.
* Enable Data Insights:
Develop tools and dashboards for data collection, analysis, and visualization, empowering the team to make data-driven decisions.
* Manage Data Flow:
Design and maintain scalable data pipelines for ingesting, processing, and versioning massive datasets from our robotics fleet.
* Optimize Performance:
Improve and maintain tooling for both on-robot and off-robot model inference, focusing on latency, throughput, and efficiency.
* Collaborate and Scale:
Partner with central infrastructure teams to optimize shared resources (e.g., compute clusters) and drive improvements that benefit the entire organization.
**** The Ideal Candidate Is...
***** A Software Pragmatist:
You are a software engineer first and foremost. You find joy in building tools, automating processes, and creating robust systems that make others more productive.
* A Force Multiplier:
You understand that great engineering is what turns brilliant ideas into reality. You are passionate about building systems that multiply the team's effectiveness, allowing them to experiment faster and more reliably. Your success is measured by the velocity and impact of the entire team.
* Committed to Quality:
You believe that testing, clean code, and solid architecture are not afterthoughts but are fundamental to moving fast and building things that last.
* A Systems Thinker:
You are comfortable working across the full stack, from data ingestion and databases to training clusters and on-device inference.
***
* Required Qualifications:

***** 6+ years of professional experience designing, building, and maintaining production Python applications.
* Proven experience deploying and optimizing neural network models in production or real-world environments.
* Deep expertise with modern software development practices: build systems (like Bazel or Pants), monorepos, Docker, and Python packaging.
* Strong familiarity with the ML ecosystem, including PyTorch, ONNX, and inference servers like NVIDIA Triton.
* Hands-on experience implementing distributed (multi-GPU, multi-node) training on a compute cluster.
* Proficiency with production-grade database systems (e.g., Postgre

SQL), ORMs, and data orchestration tools (e.g., Airflow).
**** Nice to Have:
***** Experience in robotics, behavior learning, or computer vision (VLMs).
* Familiarity with modern C++.
* Experience with front-end or web development for building internal tools (e.g., React, Vue).
* The salary or hourly pay range for this position will be clearly stated in the job posting as required by Massachusetts law. The base pay range for this position is between $– $. Base pay will depend on multiple individualized factors including, but not limited to internal equity, job related knowledge, skills and experience. This range represents a good faith estimate of compensation at the time of posting.
* As a
** Staff Software Engineer
** on the
** Atlas team**, you will be a critical engineering pillar for a world-class group of engineers and scientists creating the next generation of humanoid robotics. Our team is pushing the boundaries of Large Behavior Models, and your role is to build the robust, scalable, and efficient software foundation that accelerates our development cycles.
#J-18808-Ljbffr