Cactus Compute
Develop high-performance kernels and an AI inference engine for phone chips to enable low-latency AI on mobile devices and wearables.
At a Glance
- Mobile App Developers
- Hardware Manufacturers
- Wearable Tech Companies
AI Tools by Cactus Compute
(1)Needle
Tiny LLM for Edge Devices
Discussions
No discussions yet
Be the first to start a discussion about Cactus Compute
Latest News
Cactus Compute introduces Needle, a 26M parameter model for mobile tool-calling.
Cactus Team announces multiple papers accepted at ICLR 2026 workshops.
DeepMind x Cactus Compute Hackathon results released.
Cactus Compute joins Y Combinator S25 batch.
Products & Services
An open-source, high-performance inference engine and unified cross-platform framework for running AI models locally on mobile devices and wearables.
A specialized 26M parameter model distilled from Gemini, optimized for single-shot function calling and tool-use on mobile devices. MIT-licensed.
Market Position
Positioned as a more efficient, low-level alternative to generic mobile AI frameworks, focusing on kernel optimization and tiny, specialized models (26M parameters) like Needle.
Leadership
Founders
Roman Shemet
Former quant and economist with a background in product and data. Graduated from the University of Oxford. Experience in corporate finance, financial econometrics, and machine learning.
Henry Ndubuaku
Background in EECS, Robotics, and AI. MS in AI from the University of Pennsylvania (UPenn). Author of 4 ICLR papers and specialized in on-device AI and robotics.
Executive Team
Roman Shemet
Co-Founder
Former quant & economist, Oxford graduate.
Henry Ndubuaku
Co-Founder
Robotics and AI expert, MS AI from UPenn, 4x ICLR author.
Board of Directors
Founding Story
Cactus Compute was started to solve the latency and efficiency bottlenecks of running AI models on edge devices. The founders leveraged their backgrounds in finance (econometrics) and robotics/AI to build a specialized engine for phone chips.
Business Model
Revenue Model
Open core model with enterprise licensing (Pro Key) and potentially support/services. Free for hobbyists.
Pricing Tiers
Open-source access to models and framework for non-commercial or personal use.
Requires a 'Pro Key' and enterprise licensing for commercial deployment and support.
Target Markets
- Mobile App Developers
- Hardware Manufacturers
- Wearable Tech Companies
- On-device mobile AI assistants
- Wearable device AI features
- Function calling for mobile agents
- Private and offline AI inference
- Hobbyist developers
- Mobile AI engineers