· Research framework
A practical framework for private local AI inference
Evaluate privacy boundaries, memory pressure, latency, quality, and recovery instead of relying on a single tokens-per-second number.
Read the frameworkAI research lab
Puma AI Private local inference, agent harnesses, and memory store.
We study how useful AI can run closer to the user — practical methods for private local models, observable agent workflows, and durable memory stores.
Research notes
Methods for practical AI research
Three working frameworks for experiments we want to make measurable and reproducible.
· Evaluation method
How to evaluate an AI agent harness
Measure the system around the model: tools, state, permissions, control flow, traces, failures, and human intervention.
Read the method· Testbed design
Interactive environments as agent testbeds
Interactive scenarios offer visible state, bounded actions, repeatable setups, and failures that are easier for humans to inspect.
Read the proposalActive programs
What we are working on
Research tracks organized around privacy, control, and legible system behavior.
Private inference
Finding model, runtime, and hardware combinations that keep sensitive context close to the user.
Agent harnesses
Building evaluation loops that make tool use, state transitions, decisions, and recovery inspectable.
Memory store
Building durable memory that agents can write, retrieve, and inspect across long-running work.
Long-term vision
AI that is private, steerable, and safe
Private compute
More intelligence should run locally, with cloud services used deliberately and visibly.
Inspectable workflows
Agent systems should expose decisions, failures, permissions, and tradeoffs to their users.
Memory store
Useful agents need memory that persists beyond a single prompt and stays under user control.
Backed by builders
Angel and pre-seed support from leaders in crypto and AI.
- Stefan Thomas creator of Interledger, BitcoinJS
- Chris Larsen founder of Ripple
- Anatoly Yakovenko founder of Solana
- Illia Polosukhin founder of NEAR; co-author of “Attention is all you need”
- Sridhar Ramaswamy CEO of Snowflake; founder of Neeva
- Don Ho founder of Quantstamp, Orange DAO
- Kartik Talwar founder of ETHGlobal, A.Capital
- Jason Warner founder of poolside.ai; former CTO of GitHub
- Oleksandr Maksymets AI Research at Meta Superintelligence Labs
- Evil Rabbit founding designer & brand architect at Vercel
- and many others
Work with us
Help make local AI practical
We are interested in engineers and researchers who care about private inference, reliable agent systems, and unusually good test environments.