on-prem-ai
(4)
Inflection AI
Inflection AI is a Palo Alto-based artificial intelligence company, structured as a public benefit corporation, founded in 2022 by Mustafa Suleyman, Reid Hoffman, and Karen Simonyan. Originally known for the empathetic consumer chatbot Pi and its Inflection large language models, the company was reshaped in March 2024 after Microsoft licensed its technology for roughly $650 million and hired most of its team. Under CEO Sean White, Inflection re-founded itself as an enterprise-focused AI provider, selling data-secure, customizable models (Inflection 3.0) that let organizations own their intelligence, data, and infrastructure - including an on-premises appliance developed with Intel.
Legion Intelligence
Legion Intelligence is a San Francisco company building a governed, agentic AI platform for national security and regulated enterprises. Formerly Yurts, it lets defense agencies and Fortune 500 teams deploy and orchestrate AI agents across cloud, on-prem, air-gapped, and classified networks (IL-2 through IL-6) using any large language model, with human-in-the-loop control and full auditability. Its platform has been fielded at U.S. Special Operations Command, the U.S. Army, and the Department of Energy, where it cut intelligence-summary creation time by 94% and saved staff hundreds of hours per week.
Pokee AI
Pokee AI is a Bellevue-based startup building foundation AI agents that plan, reason, and act across thousands of internet tools. Founded in 2024 by former Meta applied-reinforcement-learning lead Zheqing (Bill) Zhu, the company uses reinforcement learning rather than pure LLM function-calling to sequence tool use reliably, claiming its foundation model outperforms GPT-4o, Claude 3.7, and Gemini 2.5 Pro on tool selection. Pokee raised a $12M seed round in 2025 led by Point72 Ventures, ships an enterprise platform with private/on-prem deployment, and open-sourced PokeeResearch-7B, a state-of-the-art small deep-research agent.
Poolside
Poolside is a San Francisco AI lab building foundation models purpose-built for software engineering. Founded in 2023 by former GitHub CTO Jason Warner and serial entrepreneur Eiso Kant, it trains its frontier models (malibu, point) with a technique called Reinforcement Learning from Code Execution Feedback - then deploys the whole stack inside customer environments for Global 2000 enterprises, financial institutions and the public sector.