WortelAI

Lahore, Pakistan · 2-10 employees · wortel.ai

Wortel AI is a data annotation and ML engineering company. In 8 years we have delivered 300+ annotation projects, labeled 400,000+ images, and annotated 30,000+ hours of egocentric video for physical AI training. Our difference: we train computer vision models for a living. We have shipped production CV systems in agriculture, smart buildings, and real estate media, so we know how bad labels break a model. Specialization: robotics and physical AI. Our largest program is egocentric video annotation for a lab training Vision-Language-Action models: temporal segmentation into atomic actions, structured left/right-hand action descriptions with camera-motion and scene tags, and QA review of other vendors' output. Methods: three-layer QA (annotator, reviewer, lead audit), inter-annotator agreement tracking, model-assisted pre-labeling where the task allows. We work in client tools or our own: [tools]. Workforce: ~200 trained annotators under full-time project leads, plus in-house ML engineers. Founder/CTO has 10 years of applied CV/ML, lead YC funded startup as ML Tech Lead / Manager and is Top Rated on Upwork. Security: NDAs with every annotator, role-based access, work confined to client-approved platforms, no data on personal devices. Clients: a physical AI lab, an agtech company, a European smart-building provider, a US real estate media company. References on request.