Tuva AI
Tuva AI, Inc. offers Rao, an always-on scientific collaborator that builds a persistent knowledge graph of a researcher's project, automatically tracking how files, ideas, and results relate. Rao integrates directly into VS Code and the command line, providing context-aware assistance for ideation, data analysis, and paper writing. The platform uses a local-first architecture, keeping research data on the user's device while routing only transient queries to third-party LLM providers.
- AI Agents
- Artificial Intelligence
- Software Only
- Data & Analytics
Funding
Founders
Founder details are not available yet.
Product
Problem
Researchers often lose time and cognitive effort rediscovering context across scattered files, notes, papers, and scripts. Standard AI assistants lack persistent project memory, forcing users to repeatedly re-explain their work and making it difficult to trust outputs or find specific results buried in their research.
Solution
Rao is an always-on scientific collaborator that builds a living knowledge graph of a user's research project, automatically tracking how files, ideas, and results relate both scientifically and structurally. It runs scientific agents on top of this context to assist with ideation, derivations, data analysis, script generation, literature search, and paper writing. Rao surfaces directly inside the researcher's existing workflow—accessible by chat in VS Code and the command line—so users can get context-aware help without changing their routines. The knowledge graph continuously evolves as the project changes, enabling Rao to connect disparate ideas and locate specific lines, figures, or results across hundreds of papers, scripts, and notes on demand.
Target Audience
Rao serves researchers at academic institutions who need an AI assistant with persistent project memory, primarily within computational fields such as computational biology, chemistry, high-energy physics, astrophysics, materials science, and machine learning.
Features
- Living scientific knowledge base that maps project context and evolves automatically as research progresses
- Chat-based access natively integrated into VS Code and the command line, with browser extension and desktop app planned
- MCP-based connectors that share project context with agents like Claude, Codex, and Cursor, reducing token waste from rediscovery
- Capabilities spanning ideation, mathematical derivation and sanity checks, paper drafting, data exploration in Python/R/Jupyter, script generation for local or HPC environments, and literature search
- Local-first architecture that stores all research data on the user's device, with only transient message routing to third-party LLM providers
- Connects to local files, Notion, and Google Drive for context ingestion across a project's existing ecosystem