Cactus
Low-latency AI engine for mobile devices & wearables
Y CombinatorSummer 2025ActiveB2BEngineering, Product and DesignArtificial IntelligenceDeveloper ToolsHard TechOpen Source
What they do
Problem
Running AI models on mobile devices and wearables often has too much latency and resource overhead for real-time, on-device use.
Solution
Cactus provides a low-latency on-device AI inference engine optimized for mobile devices and wearables.
Who it's for
Mobile and wearable app developers and device makers who need fast on-device inference.
Founders
Roman Shemet
Founder
Former quant & economist with a background in product and data engineering. Pilot, triathlete, chess enthusiast. Working on mobile inference @ Cactus.
Henry Ndubuaku
Founder
Sidestepped an AI/ML role at Nvidia to work on Cactus.
Company timeline
- Founded2025
- YC batchSummer 2025
- StatusActive
- Team size8 people
- LocationSan Francisco, CA, USA
Similar startups
Stellon Labs
Building tiny frontier AI models that run on edge devices
DeepGrove
Frontier Intelligence. On Any Device.
RunAnywhere
The default way of running on-device AI at Scale
General Instinct
Deployable Physical AI on Any Edge Device.
Reduced Energy Microsystems
Building the lowest-power silicon for embedded deep learning and…
RunLocal AI
The first AI agent for optimizing ML model inference on edge hardware