I am interested in the layer between AI capability and useful software: systems that remember context, route tools safely, stay local when possible, and have an interface people want to use.
- AI assistants & agent runtimes — memory, tool routing, reliable execution
- Local-first software — private by default, portable by design
- Developer tooling — documentation, architecture visualization, automation
- Product-grade web apps — strong interaction design, not just demos
Building in public from India. I care about the details that make a tool trustworthy: clear state, fast feedback, and software that actually ships.
|
Local-first AI assistant runtime with graph memory, semantic tool routing, markdown-governed tools, and FastAPI + Next.js surfaces. Python FastAPI Next.js AI agents
|
A local-first context layer for the web: capture pages and AI conversations, distill durable memory, and continue work across sessions. TypeScript Local-first AI memory
|
|
A CLI for writing and maintaining useful agent documentation inside a codebase. CLI Developer tools Documentation
|
An agent skill for producing beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams. AI tooling Diagrams HTML
|
|
A personal AI assistant with voice, memory, Telegram, desktop tools, and project monitoring. Python LLMs Automation
|
An anime discovery app for browsing shows, exploring details, and tracking favourites in a polished interface. Next.js React Product UI
|
Core Python · TypeScript · JavaScript · Kotlin
AI FastAPI · OpenAI-compatible APIs · LangChain · semantic retrieval
Web Next.js · React · Tailwind CSS
Data PostgreSQL/Supabase · MongoDB · Firebase
Workflow Git · Docker · local-first architecture · automation
- Making AI assistants more reliable, inspectable, and useful in real workflows.
- Exploring durable memory and context systems for both people and agents.
- Turning developer workflows into tools that are easier to understand and maintain.