Roboflow

Roboflow provides a computer vision platform for developers and enterprises building applications that interpret images, video, and live camera streams. Its tools cover dataset management, image labeling, model training, and deployment. Roboflow Workflows connects models with application logic and downstream systems, while Inference runs models in the cloud or on users’ own hardware. Roboflow Universe supplies community datasets and models. Applications include manufacturing inspection and logistics: BNSF uses its vision technology for yard inventory and train wheel inspections.

Founded in 2019 by Brad Dwyer and Joseph Nelson, now CEO, Roboflow launched in 2020 after the pair worked on an augmented reality Sudoku game. Its engineering contributions include RF-DETR, a family of real-time detection transformers for object detection, instance segmentation, and keypoint detection. The architecture combines a pretrained DINOv2 backbone with neural architecture search to balance accuracy and latency, and predicts objects without a separate non-maximum suppression step. Its Supervision Python library provides reusable dataset, detection-filtering, and visualization utilities.

In 2026, Roboflow reported adoption by over one million engineers and over 16,000 organizations, including more than half of the Fortune 100. The company raised $40 million in 2024 in a financing round led by GV, with participation from existing investors Craft Ventures and Y Combinator.