Portfolio

Projects

Things I've built: native Android apps, full-stack web platforms, and systems tools. Most are open source on GitHub.

A native Android expense manager designed for zero-latency everyday logging. Stores transactions locally in Room SQLite and leverages Android WorkManager to reliably synchronize records to Notion and Google Sheets whenever connectivity resumes.

About this project

Engineered with an offline-first architecture to eliminate network lag when logging daily transactions. Room SQLite handles transactional ACID consistency locally, while Kotlin StateFlow drives instantaneous UI state updates. Background synchronization runs smoothly via Android WorkManager with exponential backoff.

  • Kotlin
  • Jetpack Compose
  • Room SQLite
  • WorkManager

A centralized academic web portal built for IT students and faculty at AU Rajkot. Ingests departmental Excel lecture logs to deliver instant attendance calculations, weekly breakdown matrices, and compensation tracking.

About this project

Replaced scattered spreadsheets across the B.Tech department with an automated ingestion parser. Students and faculty can instantly look up weekly subject attendance, lecture counts, and medical compensation status through a responsive, mobile-optimized dashboard.

  • Python
  • Flask
  • JavaScript
  • REST API

A real-time competitive hackathon scoring platform that polls each team's GitHub repository every two minutes, verifies commits against milestone criteria, and broadcasts live XP updates to spectators through Server-Sent Events (SSE).

About this project

Automates hackathon judging by connecting directly to team repositories. An async polling worker inspects commit SHAs, milestone flags, and commit timestamps, awarding XP in real-time. Spectators watch a live animated leaderboard powered by Server-Sent Events (SSE) and Next.js.

  • Next.js
  • TypeScript
  • MongoDB
  • SSE

A production-oriented enterprise MVP developed for data-driven customer workflows. Couples a conversational assistant with a Retrieval-Augmented Generation (RAG) knowledge engine, automated lead capture pipelines, and an administrative control panel.

About this project

Designed as a reliable enterprise assistant combining semantic document search with conversational guidance. Documents are chunked and vectorized, enabling high-precision contextual answers while capturing customer leads into a PostgreSQL relational ledger.

  • Python
  • FastAPI
  • PostgreSQL
  • Docker

A real-time face detection pipeline built with FastAPI and MediaPipe. Detects facial landmarks and features per frame, streaming processed frames to a React frontend over low-latency MJPEG and WebSockets inside Docker.

About this project

High-speed computer vision pipeline processing camera frames at 30fps. Streams annotated frames and facial landmark coordinates over persistent binary WebSockets to a React client with zero perceptible latency.

  • FastAPI
  • MediaPipe
  • React
  • Docker

An adaptive quiz platform engineered for the Code Carnival 2 hackathon. Leverages Gemini AI to dynamically adjust question difficulty based on student response patterns, featuring streaks, leaderboards, and GATE exam sets.

About this project

Built for the Code Carnival 2 hackathon, Quizi monitors a student's answer patterns and dynamically adjusts upcoming question difficulty using Gemini API prompts, featuring streaks, scoring multipliers, and GATE CS sets.

  • TypeScript
  • React
  • Gemini API

A fast, suspense-driven interactive web game inspired by the series Money Heist. Engineered with TypeScript for strict type safety, responsive scoring logic, randomized question state, and fluid UI micro-interactions deployed on Vercel.

About this project

A zero-dependency client-side game engine built in TypeScript. Features randomized question sequences, high-pressure countdown timers, dynamic audio-visual feedback, and global edge distribution via Vercel.

  • TypeScript
  • React
  • Vercel Edge

An intelligent crop diagnosis platform. Users upload field photos of crops to receive instant disease classification, severity metrics, and recommended treatment advice, with a historical record of past inspections.

About this project

Assists agricultural workers with on-the-spot crop health diagnosis. Deep learning classifiers evaluate uploaded leaf imagery, estimate infection severity, and generate step-by-step treatment protocols.

  • Python
  • Flask
  • React
  • SQLite

A complete inventory, point-of-sale, and billing management system built for plumbing hardware businesses. Handles GST invoice generation, live stock auditing, customer credit ledgers, and supplier procurement.

About this project

Engineered for hardware wholesale and retail operations. Supports live inventory barcode auditing, GST-compliant invoice generation, customer credit ledgers with automated balance summaries, and vendor procurement.

  • FastAPI
  • React
  • PostgreSQL