Explore ZENO, my voice assistant, and SignSpeak ASL, my browser-based gesture recognition project.
AI/ML student · Building intelligent systems
I'm an AI/ML student who enjoys solving problems and turning ideas into working software. My interests span artificial intelligence, computer vision, voice automation, and full-stack development.
I build across the stack — from models and backend APIs to web interfaces and automation tools — with a focus on making technology useful in everyday life. Alongside my projects, I practice DSA and strengthen my foundations in artificial intelligence and machine learning.
I learn best by experimenting, asking questions, and refining what I build.
🎙️ ZENOContext-aware voice and work assistant with proactive briefings, local speech recognition, and secure mobile control. Python · Gemini API · FastAPI · faster-whisper |
Real-time ASL recognition in the browser, using hand landmarks, joint-angle kinematics, and on-device processing. JavaScript · MediaPipe · WebAssembly |
🛡️ VigilanceUrban road intelligence using transit fleet cameras, edge AI, spatial deduplication, and a WebGIS command center. Python · YOLOv8 · ONNX · FastAPI · Next.js |
Learn-once, replay-anywhere voice automation for Android, built for the Samsung PRISM GenAI Hackathon 3.0. Kotlin · Python · Android · Voice automation |
| Technologies | |
|---|---|
| Languages | Python · C++ · C · Java · JavaScript · SQL |
| AI & vision | PyTorch · OpenCV · MediaPipe · YOLOv8 · Scikit-learn · Gemini API |
| Backend & web | FastAPI · WebSockets · Next.js · REST APIs |
| Tools | Git · GitHub · Linux · Docker · VS Code |
A closer look at what I'm building
Gemini-powered conversation and a bounded observe–act–verify loop for supported system and file tasks. Work and system briefings bring together notifications, Git changes, reminders, and system health. Local speech recognition, paired devices, and action approvals support desktop and mobile use. Optional inbox connections provide read-only previews.
MediaPipe landmarks and WebAssembly support on-device gesture recognition. Joint angles, temporal smoothing, and motion tracking help distinguish similar hand shapes and moving gestures.
Quantized YOLOv8 detects road distress from transit fleet cameras. Spatial clustering consolidates repeated detections, while repair prioritization and a Next.js WebGIS interface help organize incidents.
Always learning. Always building. One idea at a time.
LinkedIn ·
Let's connect ·
Explore my work ↗
