
Helixis
a persistent memory layer for AI agents — recalls, pinned context, and tool history so language models can hold longer threads of attention across sessions.


Started on Java and HTML/CSS in school, then settled into the React + TypeScript stack through college projects. The substance was always in the late-night side builds, not the syllabus.
Interned at Tata Elxsi on autonomous-driving simulation — collision detection on CARLA + OpenMaps, OTA progress tracking for infotainment firmware. Then 1 year 4 months as a Software Development Engineer at White Rabbit Labs in Kochi: Angular 20 + NgRx applications with 100+ concurrent components, FastAPI microservices, OAuth2 QuickBooks integration with HMAC-secured webhooks, and a Notion-style real-time state layer that cut API calls by 80%.
Now at YE Stack in a hybrid product + engineering role, building Helixis — a persistent memory layer for AI agents — and broader systems that map how human cognition actually works.

a persistent memory layer for AI agents — recalls, pinned context, and tool history so language models can hold longer threads of attention across sessions.

an architecture that lets sites expose page-specific tools through llms.txt — browser agents invoke them directly, no DOM parsing.

a loan management app for Kudumbashree members — Kerala's women-led microfinance network. tracks group repayments, savings, and disbursal cycles.

a real-time rocket launch tracker — every mission in one place, push alerts for ISS passes, and background sync that cuts network calls by 60%.

a persistent memory layer for AI agents — recalls, pinned context, and tool history so language models can hold longer threads of attention across sessions.







a slow, ongoing series shot mostly between deploys — canals in Fort Kochi, tea estates in Munnar, monsoon afternoons. I post one or two a month to glass.photo.

I build my own 5-inch FPV drones from parts — soldering ESCs, flashing PX4, dialling in PID. flying is the least of it; the joy is in the build, the crash, the rebuild.
the kit I trust day-to-day, and a smaller pile I'm still learning.
three teams, four years of school, and a fair amount of overlap between the two.
building Helixis, a persistent memory layer for AI agents, plus broader systems that try to map how human cognition actually works. wear product and engineering hats roughly equally.
Angular 20 + NgRx apps with 100+ concurrent components, FastAPI microservices, OAuth2 QuickBooks integration with HMAC-secured webhooks, Notion-style real-time state that cut API calls by 80%.
collision-detection algorithm on CARLA + OpenMaps for autonomous driving research, plus an OTA progress tracker for infotainment firmware with rollback support.
coursework in DSA, DBMS, networking, web dev, and basics of neural networks. did most of my serious projects outside the syllabus.
best fits: AI agent tooling, memory / context systems, FastAPI or Angular product work, and small research-led builds. day job is at YE Stack, so I take on side work selectively.