My Work
Here are some of the projects I'm proud to have worked on.
Verilabs: In-Browser Image Labeling Tool for Automated QA
Designed and developed Verilabs, a free, in-browser tool for image labeling in object detection, aimed at enabling more flexible and private industrial quality inspection systems.
Reqlify: Requirements Management for Regulated Industries
Designed and developed Reqlify, a specialized web application to streamline requirements management for small to mid-sized companies and their numerous suppliers in regulated industries, addressing fragmented communication and costly delays.
Comprehensive NLP Model & RAG System Evaluation
A deep dive into evaluating Natural Language Processing (NLP) models, with a specialized focus on Retrieval-Augmented Generation (RAG) systems. Explores foundational metrics, common failure modes, advanced evaluation frameworks, and strategies to overcome the data bottleneck.
DocInsights: Natural Language Document Search Platform
Developed DocInsights, a comprehensive web platform enabling users to 'understand' and search hundreds of their documents using natural language, powered by Retrieval-Augmented Generation (RAG) and a scalable microservices architecture on Google Cloud.
RAGCore: Simplify Retrieval-Augmented Generation (RAG) Apps
Developed RAGCore, an open-source Python library that simplifies building Retrieval-Augmented Generation (RAG) applications, allowing developers to create robust RAG systems with just a config file and four core methods.
Neural Style Transfer Web App (NeuralCeption)
Developed an end-to-end deep learning project to deploy Neural Style Transfer models directly in the browser, addressing significant resource constraints for client-side execution.
Driver Assistant iOS App
Developed an end-to-end iOS application featuring on-device object detection for traffic elements (pedestrians, traffic lights, stop signs) and real-time speed calculation, optimized for mobile performance.
COCO Traffic Dataset: Extension & Refinement
Created and curated COCO Traffic, an extension of the COCO dataset, by relabeling over 10,000 traffic light annotations into specific states (red, green, N/A) and extending with external data, significantly enhancing its utility for traffic-related object detection. Featured on the official COCO dataset website.
Basics of Object Detection with CNNs
A comprehensive project exploring the evolution, key algorithms (Faster R-CNN, YOLO, DETR), and essential evaluation metrics of object detection in computer vision.
Adversarial Examples in Computer Vision
Explored how subtle, imperceptible image changes can mislead machine learning models, and developed practical attack implementations to understand model vulnerabilities.