Senior ML Engineer, Computer Vision

About The Job:

Berkshire Grey is a leader in the field of AI and robotics, providing innovative solutions for e-commerce, retail replenishment, and logistics. Our technology automates complex pick, pack, sort, and unload operations.

As part of the Scoop engineering team, you will develop new approaches to solving challenging trailer-unload computer vision problems for real-world robotic systems. Your work will help robots better perceive, reason about, and interact with dynamic trailer environments, including varied package types, shifting walls, confined spaces, and complex unload conditions. These efforts will enhance the performance, reliability, and throughput of our robotic unloading solutions while unlocking new value for customers. This position offers a unique opportunity to work at the cutting edge of robotics applied to one of the most demanding real-world logistics challenges.

Responsibilities:

Develop solution prototypes for computer vision problems, to improve our robots’ ability to solve increasingly complex tasks at unprecedented speeds

Quickly prototype solutions, creating demos for stakeholders and visitors

Serve as subject matter for transitioning prototypes to product teams

Identify high impact areas for improvements of our robotic systems to solve real problems

Stay abreast of the latest advancements in robotics and related fields, evaluating applicability to our challenges

Assist with mentorship of more junior engineers or interns

Communicate technical priorities and status.

Minimum Qualifications:

Master’s degree in Robotics, Machine Learning, Computer Vision, Computer Science or a closely related field.

4+ years of experience in software development with a focus on robotic manipulation or related areas.

Strong development expertise in Python and C++.

Experience with major deep learning frameworks such as PyTorch.

Experience with data science tools & libraries like numpy, pandas, scipy, matplotlib, scikit-learn

Demonstrated experience training and adapting of existing machine learning architectures for computer vision, such as CNNs/ViTs or VLMs, to solve tasks such as grasp estimation, object detection/segmentation, depth estimation, or anomaly/outlier detection

Demonstrated proficiency to solve real world computer vision problems with machine learning

Demonstrated ability to:

Develop on and troubleshoot real robotic systems

Determine and communicate justification of technical priorities

Rapidly prototype and iterate on solutions to challenging problems

Work independently on a variety of projects while maintaining focus

Work in a fast-paced environment with changing priorities

Mentor junior engineers

Strong verbal and written communication skills, capable of explaining complex ideas clearly and concisely to both technical and non-technical stakeholders

Provide technical leadership on key projects

Preferred Qualifications:

MS / PhD in Machine Learning, Computer Vision, Computer Science or a closely related field.

Demonstrated technical proficiency in computer vision applied to robotic manipulation, such as grasp estimation or long-tailed object detection & segmentation in clutter

Experience with:

Robotic vision sensors and camera to robot calibration

Both RGB and depth data

Collecting and training on real and synthetic datasets, including various forms of data augmentation

Applying machine learning to hardware interacting with the real world

Real-time perception-based control

Robot simulators (e.g. Isaac Sim)

Combining model-based and data-driven approaches

Docker, cloud computing, or similar applications

Experiment tracking and dataset management (e.g. Weights & Biases)

Database systems as data source, such as MongoDB

Parallel/distributed systems and asynchronous/concurrent programming

ROS or ROS2

Possessing some of these preferred qualifications is great to have but not required and should not discourage applicants from applying.

6110-2607RS