Multimodal lakehouse for ingesting, curating, and training on unstructured data
Ocular AI is a YC W24 startup building infrastructure for multimodal AI training pipelines. The tech stack (PyTorch, TensorFlow, OpenCV, PostgreSQL, AWS) and active projects (model training infrastructure, data pipelines for evaluation, annotation platform) show a focus on the full data-to-model lifecycle. Pain points cluster around early sales motion and closing enterprise deals, while the senior-heavy hiring mix suggests they're assembling core technical leadership to scale.
Ocular AI provides a multimodal lakehouse platform designed to handle video, images, audio, and other unstructured enterprise data. The product surfaces capabilities for intelligent search (natural-language queries across petabytes of video), auto-labeling via AI data agents, data lineage and versioning, and GPU-powered model training in a single interface. The company serves teams building computer vision, robotics perception, and domain-specific generative AI systems. Early stage (founded 2024, 2–10 employees), based in San Francisco, currently focused on closing initial enterprise customers.
Core infrastructure: PostgreSQL, AWS, Python, PyTorch, TensorFlow, OpenCV. Frontend: Next.js, React, Figma. Analytics/ops: PostHog, Salesforce, HubSpot, LinkedIn Sales Navigator.
Active projects include foundational AI models, model training infrastructure for LoRAs, data pipelines for evaluation, an annotation platform, and a multimodal collaborative canvas for data curation.
Other companies in the same industry, closest in size