Roboflow

Roboflow, Inc. is a computer vision platform, founded in 2019 by Joseph Nelson and Brad Dwyer, that covers the full workflow for building custom vision models: collecting and labeling image and video data, training models, and deploying them to production. Its annotation tool supports AI-assisted labeling for bounding boxes, polygons, and keypoints, while its training component offers cloud-hosted, no-code training with GPU access and version tracking across model iterations. A low-code "Workflows" interface lets teams assemble multi-step vision pipelines without writing custom code for every step. Once a model is trained, Roboflow supports several deployment paths, including a serverless hosted API, on-device and edge deployment, self-hosting, and integration into existing software through SDKs. The platform also runs "Universe," a public library of open-source datasets and pre-trained models that other users can search and build on. Roboflow integrates with major cloud providers, camera hardware makers, popular ML frameworks like TensorFlow and PyTorch, and tools such as Unity, Unreal Engine, and Kubernetes. Roboflow serves developers and enterprises across industries including manufacturing, healthcare, logistics, retail, and robotics, for tasks such as quality inspection, defect detection, and inventory tracking. The company states it is used by over a million engineers and more than half of the Fortune 100, and that it is SOC 2 Type 2 compliant.