February 2022 Summaries

Tensorwerk co-founder and CEO Luca Antiga announces the company’s collaboration with Grid and the PyTorch Lightning community, describing it as a partnership grounded in a shared goal of making machine learning and software development more accessible, intuitive, and powerful. He highlights Tensorwerk’s people-centered origins and its work on challenges such as machine-learning abstraction, composition, and enterprise-grade model serving. The partnership aims to extend these capabilities beyond large organizations to a broader developer community as deep learning becomes increasingly central to data-driven software development.

Clutterbot, a New Zealand startup developing autonomous home-cleaning robots, uses computer vision and artificial intelligence to navigate homes, identify objects such as toys on floors, and sort them into containers through a mobile app. Senior Machine Learning Engineer Syed Riaz and his team faced costly delays from having to build and maintain their own machine-learning infrastructure while developing the robot’s object detection, tracking, and multitask-learning capabilities. To reduce this burden, the company adopted Grid after discovering it through PyTorch Lightning, using Grid Sessions to prototype and train models on demand with automatically allocated GPUs and preinstalled development tools. The platform’s ability to pause and resume sessions helped Clutterbot avoid unnecessary computing costs, while reducing infrastructure work, CUDA and driver setup, and other operational overhead so the team could concentrate on research, product development, and business priorities.