Agility Robotics builds Digit, the world's first commercially deployed humanoid robot — already at work in real warehouses and manufacturing facilities today. Digit has proven itself in structured work cells, and now we're pushing beyond them: into open floors, shared aisles, and dynamic spaces where people, forklifts, and other robots are all moving at once. That shift raises the bar for perception dramatically — Digit needs to reliably understand and react to environments that are no longer predictable or controlled.

That means perception isn't just about capability — it's about trust. Every person Digit works near is counting on our perception systems to see them, understand them, and never get it wrong, whether Digit is in a fixed work cell or navigating a busy, ever-changing facility. We're looking for a Senior AI Software Engineer, Perception to own the design and implementation of the algorithms that let Digit perceive its surroundings and stay safe as we take on these more dynamic, less constrained environments.

What You'll Do

• Design, develop, and deploy perception algorithms (e.g., human detection algorithm) that enable safety and efficiency, defining how Digit perceives and reasons about humans and hazards as it expands from structured work cells into open, dynamic environments
• Drive adoption of state-of-the-art perception models by evaluating emerging research and technology and deciding what's worth building into future robot architectures
• Promote best practices in model design, implementation and testing to deliver high-quality, scalable software
• Optimize deep neural networks and associated data pipelines to run efficiently on embedded systems (e.g., model structure optimization, CUDA inference etc.)
• Set standards for verification and validation of ML-based perception models, especially for rare and safety-critical scenarios
• Collaborate across navigation, controls, and hardware teams to ensure the perception model serves the whole robot

What You Bring

• 5+ years of experience in machine learning/perception for robotics
• Master's or Ph.D. in Artificial Intelligence, Robotics, Computer Science, or a related discipline, with a strong foundation in machine learning, robotics, and intelligent systems
• Demonstrated experience building ML/perception systems for safety-critical or safety-rated contexts — human detection, automotive, robotics, or a comparable regulated domain
• Deep expertise in convolutional neural networks, multi-object tracking, human detection, data association, supervised learning, and pose estimation, with strong intuition for where these approaches break down in unpredictable, real-world conditions
• Experience designing and optimizing algorithms for efficient execution across CPU and GPU architectures
• Experience with MLOps, including data annotation services, data storage, model evaluation tools, and model deployment
• Publications in the field (CVPR, ICCV, RSS, ICRA preferred)

Nice to Haves

• Experience designing and implementing perception systems for humanoid or legged robots specifically
• Experience building systems based on multiple sensor modalities including camera, lidar, infrared camera. 
• Knowledge in traditional geometric computer vision and sensor hardware etc.

This a hybrid position based out of one of our Salem, Pittsburgh, or Fremont offices.

The final salary offered to a successful candidate will be dependent on several factors that may include but are not limited to: market location, job-related knowledge, skills, and experience. This range may change based on geographical location and may be modified in the future.