Available for internship & collaboration

Piyush Das

Diploma student building AI-powered, offline-first applications.
I focus on performance engineering, multi-LLM integrations, and software that works reliably without a network.

DEV.to April Fools Developer Challenge Winner — awarded top prize for Volex (Volume by Exercise)

Featured

What I'm building

Two active flagship projects — both in development, both solving real problems I care about.

More Projects

AniRec

Shipped

Full-stack AI anime recommender & personal library. Orchestrates Gemini Flash for real-time metadata parsing and Gemini Pro for narrative recommendations across 4 external APIs.

React Node.js SQLite Gemini

CivicLens AI

Hackathon

AI-powered civic complaint platform built with React + FastAPI. Implements semantic response caching to eliminate redundant Gemini API calls and reduce latency significantly.

React FastAPI Gemini AI Python

WhatsApp Chat Viewer

Shipped

High-performance browser app for parsing and rendering exported chat logs locally. Combines DOM Virtualization and Web Workers for smooth scrolling through gigabyte-scale files.

Vanilla JS Web Workers DOM Virtualization

AI-to-Docx Converter

Shipped

Next.js utility that intercepts AI chat outputs and compiles LaTeX math equations into native Word OpenXML Math (OMML) — fully editable in Microsoft Word, no plugin needed.

Next.js TypeScript Pandoc OMML

Volex

DEV.to Winner

Fitness tracking web app that won the DEV.to April Fools Developer Challenge. Zero-dependency Vanilla JS with dynamic workout volume calculation and local IndexedDB storage.

Vanilla JS IndexedDB CSS
Highlights

Selected achievements

DEV.to April Fools Developer Challenge Winner

Awarded top prize among global submissions for Volex — a zero-dependency workout volume tracker.

Apr 2026

Built a dual-LLM recommendation pipeline

AniRec: orchestrates Gemini Flash for real-time entity extraction + Gemini Pro for narrative recommendations across 4 APIs.

Feb 2026

DOM Virtualization for gigabyte-scale logs

WhatsApp Chat Viewer: engineered viewport windowing + Web Worker threading to scroll multi-GB chat exports without memory crashes.

Apr 2026

Actively building Stash AI & GhostScreen

Two flagship projects in active development — an AI-powered knowledge manager and an Android privacy utility.

2026 →
Skills

Technical toolkit

Languages

TypeScript
JavaScript
Python
Kotlin
HTML / CSS

Frontend

React
Next.js
DOM Virtualization
Web Workers

Backend

Node.js / Express
FastAPI
REST API Design

AI

Gemini API (Pro & Flash)
Model Context Protocol
Multi-LLM Orchestration
Prompt Engineering

Storage

SQLite
IndexedDB
Offline-first design

Tools

Git / GitHub
Antigravity IDE
Android Studio
Gemini CLI
Contact

Let's connect

Open for internship opportunities, collaboration, and interesting conversations about AI and offline-first software.