Loading...
Loading...
My Projects
Invoke — Voice Coding Agents
Coding Agents
Electron / Node.js / Python
A Windows voice assistant for running coding agents. It handles simple requests itself and hands bigger coding tasks to GitHub Copilot CLI agents, each working in its own Git worktree so several can run at once. Speech-to-text, the LLM, and text-to-speech all run locally (Parakeet, quantized Gemma/Qwen on llama.cpp, and Kokoro). Speculative decoding with a draft model made generation 35% faster.
ElectronNode.jsPythonllama.cppMCPCopilot CLI
Jarvis — Voice Dispatch Agent
Voice Agent
JavaScript / Node.js
A voice AI dispatch agent for a demo logistics app. Place a call and it looks up loads, assigns carriers, negotiates freight rates, and submits quotes while you talk to it. It uses an LLM planner with MCP tools and RAG over load and carrier data, streams audio over a long-lived WebSocket, and keeps the call alive as you move between pages. Live demo at jarvis.whoisdhruv.com ↗.
Voice AIMCPRAGTool CallingWebSocketsNode.jsAudioWorklet
This Portfolio
This Website
Next.js / TypeScript
The site you're on. Its main feature is a live voice agent you can talk to and interrupt mid-sentence: full-duplex audio over a WebSocket, with 25+ tools to navigate pages, open projects, fill forms, and run terminal commands for you. There's also a RAG chatbot with 5 selectable LLMs, one of them self-hosted, a retro terminal, and a hand-drawn sketchbook look. Built with Next.js 16 and deployed as Docker images through GitHub Actions. Live at whoisdhruv.com ↗.
Next.jsReactTypeScriptGemini LiveRAGDockerGitHub Actions
Cropio
AI Cropper
Next.js / FastAPI
A privacy-conscious AI portrait cropper built to turn raw photos into polished headshots. It blends YOLO11 pose estimation with face-orientation detection to capture different angles of the face and generate multiple headshot crops, plus a precise interactive crop editor and semantic archive search so users can upload, refine, and export professional crops with full-resolution rendering in the browser. Live at cropio.whoisdhruv.com ↗.
Next.jsTypeScriptFastAPIYOLO11 PoseIndexedDBNVIDIA APIs
Fluent UI Android
Android Lib
Kotlin / Java
Microsoft's open-source Jetpack Compose component library for building Microsoft 365 Android apps. As its primary maintainer I shipped 10 releases used by Copilot, Office, Outlook, OneDrive, and Teams, and owned the library's security compliance.
KotlinJavaAndroid SDKJetpack ComposeXMLGradleMaven
Hybrid Recommender
Movie Night
Python / ML
A context-aware movie recommendation engine that integrates collaborative, content-based, and demographic filtering to solve the cold-start problem and personalize content for diverse user groups.
PythonScikit-LearnPandasSVDTensorFlow
Bloom Filter Research
Research Paper
Research / C++
Undergraduate research at IIIT Delhi's Distributed Computing Lab on Counting Bloom Filters under high contention. I built a lock-free C++ filter with relaxed synchronization that reached 3x the baseline throughput.
C++Bloom FiltersConcurrencyOptimizationData Structures
IVC - Vital Checkup
Vitals Scan
Python / OpenCV
A contactless, computer-vision powered health screening kiosk that automates patient triage. Using OpenCV, it calculates height, weight, BMI, and pulse from a distance, drastically reducing wait times.
PythonOpenCVComputer VisionHealthTechReal-time Processing
Course Evaluator
Overlap Detector
Python
A Python tool that finds overlapping course content across university curriculums. It uses fuzzy matching and text similarity to flag modules that cover the same material, so students don't end up taking the same course twice.
PythonFuzzy LogicNLPData AnalysisAlgorithm Design
AtomVault
Bank Vault
Java / MySQL
A secure, ACID-compliant banking database built for high-reliability transactions. Features multi-user architecture with strict role-based security.
JavaMySQLJDBCOOPACID Compliance