Our Thesis
Our Thesis
OnDeck is a visual intelligence research company.
OnDeck is a visual intelligence research company.
[THE PROBLEM]
Visual perception provides the most information-dense signals for humans and machines to navigate the world.
Close your eyes and try to get a task done - you soon understand just how much information we get from sight. After decades of computer vision research breakthroughs, vision AI has yet to reach human level performance - even after LLMs have surpassed human level performance in many domains.
Why? To understand diverse and dynamic visual data like humans, visual AI needs to gain the ability to adapt like the human mind.
Visual perception provides the most information-dense signals for humans and machines to navigate the world.
Close your eyes and try to get a task done - you soon understand just how much information we get from sight. After decades of computer vision research breakthroughs, vision AI has yet to reach human level performance - even after LLMs have surpassed human level performance in many domains.
Why? To understand diverse and dynamic visual data like humans, visual AI needs to gain the ability to adapt like the human mind.
[THE SOLUTION]
We are building recursive self-improving mechanisms for vision.
This has two components: [1] powerful visual analysis primitives & tools, and [2] a system that autonomously builds a fully formed expert 'model' for any domain.
The result? OnDeck's Perception-0 is the most powerful vision model on earth.
We are building recursive self-improving mechanisms for vision.
This has two components: [1] powerful visual analysis primitives & tools, and [2] a system that autonomously builds a fully formed expert 'model' for any domain.
The result? OnDeck's Perception-0 is the most powerful vision model on earth.
Perception-0
Gemini 3.1 Pro
Deep Video Discovery
Seed1.5
AdaReTaKe
GPT-4o
Conveniently, Perception-0 is not a model in the traditional sense. It is a harness that gets remarkably better performance out of frontier models and autonomously self-improves on your task. This means as the frontier models improve, OnDeck will always provide better performance, more efficiently.
Conveniently, Perception-0 is not a model in the traditional sense. It is a harness that gets remarkably better performance out of frontier models and autonomously self-improves on your task. This means as the frontier models improve, OnDeck will always provide better performance, more efficiently.
[THE FUTURE]
As society automates, everything will need visual intelligence. We’re creating a self-improving visual reasoning engine that can solve any visual task, as well or better than a human in any domain, autonomously. This puts us on a path to be the last vision research company.
As society automates, everything will need visual intelligence. We’re creating a self-improving visual reasoning engine that can solve any visual task, as well or better than a human in any domain, autonomously. This puts us on a path to be the last vision research company.