ML data curation platform for computer vision model improvement
Lightly builds tools to curate and prepare vision data for machine learning training. The stack—Python, TensorFlow, PyTorch, scikit-learn—points to a research-forward, ML-native engineering culture. Active hiring across engineering, data, and ops roles suggests they're scaling both product development and go-to-market in parallel; the open-source Python package for self-supervised learning signals a developer-community play alongside commercial offerings.
Lightly helps machine learning teams improve model performance by automating the selection and labeling of high-value vision data. Founded in 2019 and based in Zurich, the company operates as an 11–50 person team focused on reducing the burden of preparing large, unlabeled image datasets. The product targets enterprises and research groups building computer vision systems where data volume is a bottleneck. Current hiring spans engineering, data science, and commercial roles across Switzerland, the United States, and Japan.
Lightly's core stack includes Python, TensorFlow, PyTorch, and scikit-learn for model development. The platform runs on AWS and GCP, with Docker and Kubernetes for deployment. Frontend uses React and TypeScript.
Active projects include evaluating AI-generated forecasts (transport and energy), developing self-supervised learning tools (open-source Python package), and building data labeling and curation workflows for ML training pipelines.
Other companies in the same industry, closest in size