Offline domain experts with a benchmark and a playbook.
Patent Strategist leads the pattern: a domain model, a test set, a local run path, and a use-case playbook. Security, legal, finance, and medical packs follow the same shape as they mature.
The flagship pack
Patent Strategist
An AI built to reason about patents, fully offline. The pack pairs the model with its public benchmark, so you can see how it scores before your ideas and filings ever leave the room. Built with NVIDIA NeMo, shipped as GGUF and adapter artifacts.
What a pack includes
Model
An offline model or adapter selected for one domain, not a generic chatbot wrapper.
Benchmark
A test set or scorecard that makes improvement visible before the model is trusted.
Run path
A local setup path for customer-controlled hardware and private documents.
Playbook
Use cases, prompts, limits, and review rules for the domain.
The library behind the next packs
All models →
Security
SecurityLLM
Model shipped · pack planned
Legal
Saul 7B Instruct
Model shipped · pack planned
Finance
Finance Chat
Model shipped · pack planned
Medicine
II-Medical 8B
Model shipped · pack planned
Space math
Kepler
Model shipped · pack planned
These expert models are implementation aids, not professional advice. Legal, medical, finance, and security outputs need qualified human review before use in real decisions.
Built on the NVIDIA stack
Orionfold is a member of the NVIDIA Inception program. Patent Strategist is built with NVIDIA NeMo, and packs are tested on NVIDIA DGX Spark-class hardware so they run offline on customer-controlled machines.
See the DGX Spark field lab →© 2025 NVIDIA, the NVIDIA logo, NVIDIA DGX Spark, NVIDIA Inception, NVIDIA NeMo, NVIDIA NIM, and TensorRT-LLM are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries.
