Technical Portfolio.
Deploying intelligence at the edge. A collection of research-driven projects focusing on on-device ML and architectural innovation.
Gemini Token Confidence Explorer
Visualize and analyze AI text generation with real-time confidence scores for each token using Gemini Nano. Explore the inner workings of LLM output probability.
Inference
On-Device
Model
Gemini Nano
Privacy
100% Local
On-Device Mindmap Generator
Generate interactive mindmaps from your documents using Gemini Nano entirely on-device in Chrome. Privacy-first document analysis.
Context
Document-Level
Engine
Chrome AI
Privacy
100% Local
On-Device Audio/Video Analysis
AI chatbot for analyzing audio and video content using Gemini Nano in Chrome. Process multimedia entirely on your device.
Media
Audio/Video
Engine
Gemini Nano
Privacy
On-Device
On-Device RAG AI Chatbot
Retrieval-Augmented Generation chatbot using Gemini Nano in Chrome. Chat with your documents without sending data to the cloud.
Architecture
RAG
Model
Gemini Nano
Privacy
100% Local
Restaurant Review Sentiment Analysis
Real-time sentiment classification of restaurant reviews running entirely in the browser using TensorFlow.js. Optimized for mobile.
Model
LSTM
Size
< 5MB
Target
Edge Browsers
Webcam Image Classification
Real-time image classification using your webcam with TensorFlow.js. On-device inference with MobileNet architecture.
Model
MobileNet
FPS
Real-time
Input
Webcam
Image Classification
Upload and classify images using TensorFlow.js in the browser. On-device image recognition with pre-trained models.
Model
MobileNet
Inference
Browser
Size
< 10MB
Linear Regression Visualizer
Interactive visualization of model training for line fitting using TensorFlow.js and p5.js. Real-time gradient descent exploration.
Optimizer
SGD
Visual
p5.js
Frame
60 FPS
Synthia: AI-Powered Avatar Assistant
Your personal AI avatar assistant powered by advanced language models. Interactive conversational AI with visual representation.
Type
Avatar AI
Interface
Visual
Model
LLM