Computer vision platform for workplace safety and operations at industrial sites
Voxel builds a camera-based safety platform for industrial operations using PyTorch, TensorFlow, and Kafka to process petabyte-scale video streams. The stack reveals a company in heavy ML-to-production mode: they're adopting Weights & Biases and MLflow (experiment tracking), running distributed inference at scale (TensorRT, ONNX, on-device models), and orchestrating multi-site deployments via Kubernetes and Flink. Hiring skews senior (4 of 7 roles) and engineering-heavy (5 of 7), signaling they're scaling data pipelines and inference infrastructure rather than sales breadth.
Voxel operates a site intelligence platform that uses computer vision to detect and prevent workplace injuries and safety risks at industrial and manufacturing sites. The product targets safety and operations leaders at Fortune 500 companies across grocers, retailers, manufacturers, food and beverage warehousers, and logistics providers. The company ingests continuous camera feeds from multiple facilities, runs ML inference pipelines to surface actionable insights, and helps customers reduce workers' compensation costs and liability exposure. Founded in 2020 and based in San Francisco, Voxel operates with a globally distributed team.
Voxel uses PyTorch and TensorFlow for model training, TensorRT and ONNX for inference optimization, Kafka and Apache Flink for streaming data pipelines, Kubernetes and AWS EKS for deployment, and Salesforce for sales operations. They're adopting Weights & Biases and MLflow for experiment tracking.
Active projects include petabyte-scale camera data pipelines, continuous time-and-motion studies, automated data collection methods, productized insights platforms, and on-device inference optimization for multi-site deployments.
Voxel's technology stack, projects, and hiring signals are inferred from public hiring and company data — career pages, public listings, and company web presence — then clustered and de-duplicated. Figures are estimates that refresh over time. Read our full methodology →
This is not an official vendor or customer list. It is a technology-adoption signal inferred from public data, intended for B2B research.