Mumbai, India

Building AI tools
for developers

I make codebases understandable, agent skills reliable, and LLM costs controllable. Open-source by default, commercial when it matters.

view my work hire me →
products
01 Knowledge graph
Loom

Unified code + docs knowledge graph. Semantic edges between symbols and specs. Ask "what breaks if I change this?"

loom ↗ 02 Agent skills CLI
Toolmark

Build, test, and publish AI agent skills across Claude Code, Cursor, and Windsurf. LLM-as-judge + Ed25519 signing.

toolmark ↗ 03 LLM FinOps
TokenGuard

Hard budget limits, auto model downgrade, complexity routing. Know what your LLM calls actually cost in real time.

tokenguard ↗
consulting

AI tooling for engineering teams

I help seed and Series A teams ship AI features fast — without the infra debt. Specialising in LLM integration, MCP servers, RAG pipelines, and code intelligence systems.

  • LLM integration & architecture review
  • MCP server design and implementation
  • RAG pipelines and code intelligence systems
  • AI cost auditing and model routing strategy
Available for projects · $90 / hr USD
devashi.sh
writing
How I built a code knowledge graph with FalkorDB and tree-sitter Mar 2026 Why I chose FastMCP over a traditional API for Loom Feb 2026 Ed25519 signing for AI agent skills — a primer Feb 2026 All posts →