Tirth Kanani
Building infrastructure and tools that make frontier AI systems more reliable, interpretable, and efficient in the real world.
At a Glance
- Software Engineers
- AI Agent Developers
- UX Designers
- Engineering Teams using Claude Code/Cursor
AI Tools by Tirth Kanani
(1)
code-review-graph
Local Code Graph for AI Assistants
Discussions
No discussions yet
Be the first to start a discussion about Tirth Kanani
Latest News
Products & Services
A persistent, incremental code knowledge graph purpose-built for LLM agents. Features Tree-sitter AST parsing, multi-hop dependency reasoning, and SHA-256 hashed incremental indexing. Achieves up to 49x average context compression.
A real-time AI UX team that audits screens using Vision Language Models (VLM) and provides spoken feedback from six expert personas.
An open-source tool for making sense of unstructured documents using knowledge graphs.
Market Position
Provides a unique 'structural memory layer' that outperforms basic RAG and full-repo loading by providing multi-hop dependency analysis and massive token reduction.
Leadership
Founders
Tirth Kanani
AI & Automation Specialist at Telana; Previously Graduate Research Assistant at University of Birmingham's HCI & AI Lab; MSc from University of Birmingham; Winner of Epiminds Multi-Agent Hackathon 2025.
Executive Team
Tirth Kanani
Founder
Specialist in building infrastructure and tools for reliable, interpretable, and efficient AI systems. Experienced in Python, TypeScript, and real-time Gemini Live streaming.
Founding Story
Tirth Kanani created code-review-graph to prevent LLM agents from re-reading entire codebases on every prompt, significantly reducing costs and improving review quality through a persistent structural memory layer.
Business Model
Revenue Model
Freemium subscription model for CrumbleUX; Open-source (MIT) for code-review-graph.
Pricing Tiers
Balanced audit covering hierarchy, navigation, and information architecture.
In-depth audits focusing on pixel-perfect spacing, typography, and conversion metrics.
MIT licensed open-source software.
Target Markets
- Software Engineers
- AI Agent Developers
- UX Designers
- Engineering Teams using Claude Code/Cursor
- AI code review and intelligence
- LLM context compression
- Real-time UX/UI auditing
- Structural codebase analysis
- Claude Code users
- Cursor users
- GitHub Open Source community