I build tools that give AI agents memory — and use AI agents to build faster than one person reasonably should.
These days that means an SDE role at ZySec AI (building a multi-tenant AI SOC/SOAR platform end to end), open source on nights and weekends, and the occasional script that exists purely because it annoyed me that it didn't.
Smriti — memory for AI coding assistants
Sanskrit: स्मृति — "memory, the faculty of remembrance." Your AI pair programmer forgets everything between sessions; Smriti fixes that. Persistent, cross-session memory with hybrid relevance scoring — vector embeddings, recency decay, importance weighting — plus an MCP server, React dashboard, CLI, and cross-IDE support. 330 tests passing.
VibeKit — AI-powered project management, SaaS + CLI
Sprints, tasks, and project config that stay in sync between a cloud dashboard and your local dev environment over real-time SSE. Bidirectional GitHub integration, AI task editing with intent detection, CLI published on PyPI.
siri_spotify_link — Siri ↔ Spotify on macOS, born of AirPods frustration · LipBuddy — lip reading with deep learning (a LipNet implementation)
AI-native, unapologetically. Claude Code and Cursor are load-bearing parts of my toolchain — it's how one engineer ships a security platform, an OSS project, and freelance client work in parallel. The problem that interests me most: AI can already write code; the scaffolding that makes it reliable — memory, context, tooling — mostly doesn't exist yet. That's the gap I like building in.
also fluent in: Redis · Neo4j · Helm · AWS · LangGraph · Bun · whatever the problem needs
- Winner, Smart India Hackathon 2022 (top 2% of 44,000+ teams) · Finalist, SIH 2023
- Amazon ML Summer School mentee (2023 & 2024)
- Claude Certified Architect – Foundations (Anthropic, 2026)
Gamer (it's problem-solving with better graphics), traveler (collecting experiences, not things), and permanently three tabs deep into some new rabbit hole.



