OnDeck AI

Executive Summary

OnDeck AI is a YC S25 pre-seed startup (Vancouver, 5–8 people, ~$500K raised) building a no-training-data video analytics engine using Vision-Language Models, with stated customers in fisheries, conservation, robotics, and national defense. The core technical thesis is credible — validated by a NeurIPS workshop paper and a genuine market timing tailwind as VLMs mature — and the founding team has unusual interdisciplinary depth for a company this early. The single biggest risk is a structural one: the market they're targeting (defense, enterprise video analytics) demands compliance infrastructure, long sales cycles, and capital that a $500K-funded team simply cannot sustain, while well-funded competitors like Twelve Labs ($107M raised) are converging on the same zero-shot video understanding thesis. Investors should treat this as a high-conviction technical bet on a team that needs immediate follow-on capital and a defined beachhead to avoid diffusion across too many verticals.