Suryadev
full-stack & ai/ml developer

Building systems that actually work end to end.

I'm Suryadev — I build full-stack products, wire up AI/ML pipelines, and ship Chrome extensions that inject deep into real browser workflows. Demos that run on real data, not screenshots.

based in india github ↗ linkedin ↗
scroll

Two products, built to be genuinely finished.

42% 61%
tech cast — hackathon build

Squash Gateway

An enterprise platform that cuts AI API costs by compressing prompts before they reach the model — live, inside the chat window.

4layer safety pipeline
3providers routed
MV3Chrome extension
  • Compression engine on Microsoft's LLMLingua-2 (XLM-RoBERTa-Large), with a frequency-based fallback
  • Safety net: length guard, dynamic ratio scaling, neural compression, and a context handbrake that reverts if negation words get dropped
  • PII redaction with round-trip restoration and tiktoken-based token accounting
  • MutationObserver-based composer detection with a one-tap keyboard shortcut
ReactViteTailwind FastAPIChrome ExtensionLLMLingua-2
PASS ALERT
demo build

JOURNEY by NEXA

A railway passenger-experience platform — staff and admin tooling built on top of a digital pass system, live station intelligence, and emergency handling.

25passengers seeded
10stations modeled
2roles: staff & admin
  • Separate React/TypeScript frontend and Python FastAPI backend
  • Nexa Pass digital ticketing, OTP and Google sign-in flows
  • Live station crowd data and emergency request handling
  • Backend seeded with 2 trains and 2 routes across 10 stations
ReactTypeScriptFastAPIPython

What I reach for.

frontend

  • React
  • TypeScript
  • Vite
  • Tailwind CSS

backend

  • Python
  • FastAPI

ai / ml

  • Prompt compression (LLMLingua-2)
  • Multi-provider LLM routing
  • Token-aware pipelines

extensions

  • Chrome Extensions (MV3)
  • DOM injection & observers

Open to internships, collaborations, and hard problems.