Projects

Projects by Priyanshu Chawda

A focused collection of AI agents, developer tools, and full-stack applications.

GemmaControl

Private on-device WhatsApp AI agent

Native Android AI agent that turns WhatsApp notifications into a local actionable inbox with offline FunctionGemma tool-call suggestions.

Problem: Busy WhatsApp conversations create scattered follow-ups, priority messages, and reminders, while cloud assistants introduce privacy risk for personal or work chats.

Solution: I built a Kotlin Android app that captures notifications with user permission, parses WhatsApp message context, stores sensitive data locally with AES-GCM encryption, and uses on-device FunctionGemma through LiteRT-LM to propose user-confirmed actions.

Tech: Kotlin, Android 16, Jetpack Compose, Room SQLite, WorkManager, LiteRT-LM, FunctionGemma 270M, AES-GCM, Android Keystore, MVVM

GitHub repoNative Android appTechnical documentationNotification parser

Impact: Demonstrates a privacy-first mobile AI agent pattern where notification ingestion, task extraction, reminders, priority flags, and reply preparation stay on device unless the user explicitly acts.

View case study

Screen Recorder

Native Windows capture and encoding app

Stable C++20 Windows screen recorder with WGC capture, D3D11, hardware-first H.264, WASAPI audio, camera overlay, HQ mode, and diagnostics.

Problem: Reliable Windows screen recording needs low overhead capture, synchronized audio/video output, useful diagnostics, and graceful fallbacks across different GPU and power states.

Solution: I built a native Win32/C++ recorder that captures the full screen through Windows Graphics Capture, encodes MP4 output with Media Foundation and Intel Quick Sync when available, records system audio through WASAPI, and writes per-session diagnostics beside each recording.

Tech: C++20, Win32, Windows Graphics Capture, D3D11, Media Foundation, Intel Quick Sync, WASAPI, CMake, CPack, GitHub Actions

GitHub repoLive documentation siteNative Windows appCMake build

Impact: Ships a lightweight native recording pipeline with default low-resource capture, optional 1080p HQ output, camera overlay, audio controls, orphaned partial-file recovery, and a public Netlify documentation site.

View case study

WorkAudit AI

Local-first desktop work evidence system

Privacy-first Windows work recorder for local evidence timelines and cited AI reports.

Problem: Developers, students, freelancers, and remote knowledge workers often need to reconstruct what they worked on, but manual notes lose context and many AI summaries cannot cite trustworthy session evidence.

Solution: I built a local-first Tauri + Python sidecar app that records allowed desktop activity evidence, builds deterministic local timelines, previews and deletes sensitive artifacts, and generates reports only from cited session evidence.

Tech: Tauri v2, React, TypeScript, Python 3.13, FastAPI, Rust, SQLite WAL, Local Ollama-compatible models, Vitest, Pytest

GitHub repoTauri desktop appFastAPI sidecarSQLite WAL storage

Impact: Proves a private-beta Windows workflow for first-run privacy onboarding, session recording controls, active-window and screenshot metadata capture, evidence search, share-safe exports, provider provenance, and local validation gates.

View case study

CodeAudit MCP

AI code audit MCP server

Read-only MCP server and skill for AI code audits and repository reviews.

Problem: Developers using AI agents need a safer way to inspect projects, improve code quality, review security risks, and plan issues or PRs before changes are made.

Solution: Exposes read-only MCP tools and a CodeAudit skill for project detection, skill routing, repository audits, documentation evidence checks, security review, and issue or PR planning.

Tech: TypeScript, Model Context Protocol, MCP, Node.js, npm, Streamable HTTP

npm packageskills.sh skillGitHub repoMCP server

Impact: Packages a production-style code audit MCP server on npm and skills.sh so agents can inspect repositories, improve code quality, and produce evidence-backed plans without write access or remote mutation tools.

View case study

YouTube Flashcards

Gemini AI study tool

Turns YouTube lectures into interactive flashcards for faster review.

Problem: Watching long educational videos is passive and manual note-taking slows learning.

Solution: Uses Gemini AI to transform transcripts into concise Q/A flashcards with a quick review flow.

Tech: Next.js, TypeScript, Tailwind CSS, Gemini API

Live demoGitHub repoArchitecture notesScreenshots

Impact: Converts passive lecture watching into a reviewable study workflow with transcript parsing, structured Gemini output, and editable user approval before cards are trusted.

View case study

ghfind

AI GitHub issue and PR search

AI-powered search for relevant GitHub issues and PRs using natural language.

Problem: Finding the right GitHub issues or PRs via the UI is slow and often incomplete.

Solution: Uses Gemini to interpret natural language queries and surface high-signal results quickly.

Tech: Python, Streamlit, GitHub CLI, Gemini API

GitHub repoArchitecture notesCLI integrationQuery validation

Impact: Turns vague developer questions into targeted GitHub searches, reducing manual filtering through issues and PRs while preserving direct links to original discussions.

View case study

Browser4All

AI secure browser with Gemini

AI secure browser concept with real-time phishing and malware analysis.

Problem: Phishing and malware evolve faster than traditional browser warnings can catch.

Solution: Applies Gemini-driven threat analysis to flag risky pages and downloads earlier.

Tech: TypeScript, Gemini AI, Threat analysis pipeline

GitHub repoArchitecture notesThreat-model notesScreenshots

Impact: Explores a defense-in-depth browser workflow where risky URLs and downloads are analyzed before user action, with explanations designed to reduce alert fatigue.

View case study

Smart Crowd Navigator

AI crowd navigation system

AI crowd navigation system for safer exits and better flow during events.

Problem: Large events suffer from dangerous crowding and inefficient exit timing.

Solution: Uses AI analytics to suggest safer routes and timing for crowd movement.

Tech: TypeScript, AI analytics, Simulation modeling

GitHub repoArchitecture notesSimulation modelScreenshots

Impact: Models crowd routing as an operator decision-support system with density signals, route tradeoffs, and safety-first recommendations instead of generic chatbot advice.

View case study