Software Engineer · Bangalore, India
Three years at Dell on infrastructure platforms: virtualized data grids, real-time state sync, event-driven provisioning, caching architecture. This site runs the things it claims — every demo below is live, not a screenshot.
The work I'm proudest of is invisible on paper. “4,000 concurrent connections” is a number nobody can feel. “Virtualized data grid” is a phrase, not an experience.
So each claim below is attached to a system you can run in this tab — and, more usefully, one you can try to break. There are chaos switches throughout. Please use them.
house rules
Each one takes about thirty seconds to understand and rather longer to break. Start with the first.
Scroll a hundred thousand rows without the page stuttering — then switch the trick off and watch it fall apart.
Hand-written windowing inside a custom element with Shadow DOM. Dual-mode pagination, live FPS instrumentation.
Open lab“Engineered high-performance virtualized data grids using Custom Web Components (Shadow DOM), RTK Query, and dual-mode pagination, enabling smooth interaction with 10,000+ assets without UI degradation.”Dell Technologies — Inventory Management & Infrastructure Platform
Open this page twice. Move something in one window and it moves in the other — and in every other visitor's window too.
Cloudflare Durable Object room, WebSocket delta updates into Redux Toolkit. Chaos panel for reconnect and state repair.
Open lab“Built scalable real-time state synchronization using Redux Toolkit and WebSocket delta updates, supporting 4,000+ concurrent live-sync connections across multi-tenant environments.”Dell Technologies — Inventory Management & Infrastructure Platform
Ask the same question twice and watch the second answer arrive in a fraction of the time — then break the cache on purpose.
Redis with dynamic keys, configurable TTLs and dependency-aware invalidation cascades. Real timings, no simulation.
Open lab“Designed a unified Redis caching architecture with configurable TTLs, dynamic cache keys, dependency-aware and event-driven invalidation, and cache warming to reduce redundant database workloads.”Dell Technologies — Internal Full-Stack Platforms
Drop in any PDF — your own job description works well — and watch it get read, understood and turned into structured data.
Streaming DAG: parse → OCR (WASM) → chunk → extract (LLM) → schema-validate. Per-stage latency, tokens, cost, retries.
Open lab“Engineered LLM/OCR and document-processing pipelines integrating PDF/CSV processing, cloud services, storage, and external APIs, with persistent workflow state managed through PostgreSQL/Prisma.”Dell Technologies — Internal Full-Stack Platforms
A real database, running inside your browser tab, that you can type questions into.
Postgres compiled to WebAssembly. Multi-tenant schema with live row-level isolation you can toggle.
Open lab“Architected full-stack platforms using Next.js, TypeScript, Prisma and PostgreSQL, designing a 50+ entity relational data model with indexing, constraints, and multi-tenant access patterns.”Dell Technologies — Internal Full-Stack Platforms
The short version is below. The long version is written as postmortems — what was wrong, how anyone found out, and what it cost.
education
BITS Pilani
M.E., Computer Science
2021 — 2023
Government College of Engineering, Aurangabad
B.E., Computer Science
2015 — 2019