FabricAI app

● Developed a mobile-first AI app to predict fabric type from images using a PyTorch EfficientNetB0 model

● Provide real-time predictions, trained on FABRIC dataset, optimized for mobile(40% faster, 95% prediction accuracy)

● Features include real-time camera detection, shopping & eco-friendly suggestions, and multilingual support

Location-Based Alarm app

● Developed a cross-platform map-based alarm app using Flutter, allowing users to set location-based alerts and receive real-time notifications when approaching their selected destination

● Implemented Google Maps SDK and background location services to create a robust geofencing system. Optimized for performance and battery efficiency on both Android and iOS

AI Stock Predictor

● Built a full-stack AI-powered stock prediction web app using Flask, React, PyTorch, and scikit-learn on Yahoo Finance time-series data

● Engineered RESTful APIs for real-time stock data retrieval and user requests(chart toggling, Confidence Interval)

● Designed a responsive UI and scalable backend, laying the groundwork for push notifications & an AI-powered Chatbot

● Engineered real-time data pipeline with Kafka(streaming data ingestion), feature extraction, news sentiment analysis(80% reduced data latency)

● Developed a dashboard with charts, newsfeed, AI-Chatbot, data processing & model updates orchestrated by Airflow

License Plate Recognition

● Developed a real-time license plate recognition system using YOLOv4 and TensorFlow OCR, achieving 98%+ accuracy

● Enhanced recognition by replacing Tesseract with a TensorFlow CRNN for better accuracy and distorted plates

● Optimized model for smart city deployment using TensorFlow Lite for fast & efficient inference on mobile deployment

CUNY Career Link

● A proposed full-stack career portal designed to centralize exclusive referrals directly from all 25 CUNY counselors

● Designed role-based authentication and user interfaces for students and counselors using React, Node.js, and Tailwind CSS, with RESTful APIs allowing counselors to post opportunities and students to explore and filter relevant listings

Multi-Model NoSQL Query Engine

● Optimized MongoDB queries for efficient document storage & retrieval, integrating Neo4j for graphing relationships

● Designed GUI & CLI interfaces to enhance accessibility & optimized queries(reduce retrieval time by 50%)

Drug-Gene-Disease Relationship Analysis with PySpark & Hadoop(Hetionet)

● Built a scalable data processing pipeline using PySpark, enabling efficient distributed computing

● Integrated the solution with Hadoop for large-scale data storage and processing

● Conducted performance tuning to optimize data transformations and reduce processing time by 30%

Job Assistant

● Created a Job Assistant that scraps jobs from the websites based on job-roles and locations

Mimic CPU(Rust)

Reject Me If You Can(Web Game & Python-app Game)