Case Study: Plus One Robotics achieves 25% F1 improvement and faster, scalable model iteration with Aquarium Learning

Automated Logistics With Industrial Robotics

Plus One Robotics, a company that builds vision software to power robotic arms for warehouse and e-commerce fulfillment, needed to improve the picking performance of its PickOne system and scale model iteration across customers. To support continuous model deployment, catch failures early, and reduce load on its ML team, Plus One partnered with Aquarium Learning and used Aquarium’s platform for model validation, comparison, and dataset exploration.

Aquarium Learning delivered a web-based workflow that streamlined visualization, failure-surfacing, collaboration, and model comparison, letting Plus One find and fix failures, curate data, and validate retrained models faster. Using Aquarium Learning, Plus One achieved a 25 percentage-point absolute improvement in F1 for a container-decanting model, realized performance gains across other models, saved hundreds of hours versus their prior spreadsheet workflow, and shifted day-to-day data curation to systems engineers so ML engineers could focus on deep research.

Plus One Robotics

Dan Grollman

Senior Engineer