Hassan Rasheed

Data Scientist | AI/ML Engineer | Computer Vision

Machine Learning Engineer specializing in deep learning, computer vision, and AI systems. I develop intelligent solutions that bridge advanced research with practical, real-world applications.

About Me

Data Science student at Gift University with extensive hands-on experience in Machine Learning, Deep Learning, and Computer Vision. Specialized in developing end-to-end AI solutions including foundation models, knowledge distillation frameworks, and real-time computer vision systems. Experienced in medical imaging, intelligent transportation systems, security surveillance, and healthcare analytics. Proficient in Python, PyTorch, TensorFlow, with strong skills in deploying AI applications using Flask, FastAPI, and Streamlit. Passionate about solving real-world problems through innovative AI research and development.

Education

BS Data Science

Gift University, Gujranwala, Pakistan | Dec 2022 – July 2026

Work Experience

June 2025 – November 2025

AI & ML Intern

Zapta Technologies - Lahore, Pakistan

Developed foundation model-enhanced COVID-19 detection system achieving 99.3% accuracy using knowledge distillation with BiomedCLIP, DINOv2, and OpenAI CLIP.
Engineered lightweight student model with 297x parameter reduction maintaining 96.4% accuracy through multi-level knowledge distillation.
Built brain tumor segmentation system using Attention U-Net architecture achieving Dice score of 0.993 on BRATS 2020 dataset.
Implemented medical-optimized efficient blocks with SE attention mechanisms for real-time mobile deployment (45ms inference).

May 2024 – December 2024

ML Intern

Nexl Tech - Gujranwala, Pakistan

Developed real-time traffic monitoring system using YOLOv12 and ByteTrack for red light violation detection with HSV color space analysis.
Built AI-powered construction safety monitoring system for PPE violation detection using YOLOv8 with bilingual audio feedback.
Implemented real-time weapon detection system using YOLO/SSD-MobileNet for security surveillance applications.
Created hand gesture recognition system using MediaPipe achieving 95%+ accuracy with 7 gesture classifications and 30+ FPS performance.

My Skills & Tech Stack

Programming Languages

Python

C++

Java

SQL

JavaScript

Machine Learning & AI

TensorFlow

PyTorch

Keras

scikit-learn

XGBoost

🤗

Hugging Face

Computer Vision

OpenCV

MediaPipe

YOLO

YOLO (v5/v8/v12)

Vision

Vision Transformers

U-Net

ResNet

Data Science & Visualization

Pandas

NumPy

Matplotlib

Seaborn

Plotly

Tableau

Power

Power BI

Databases

MySQL

PostgreSQL

MongoDB

Tools & Platforms

Git/GitHub

Docker

Jupyter

Google

Google Colab

Projects

Foundation Model-Enhanced COVID-19 Detection

Novel knowledge distillation framework using foundation models (BiomedCLIP, DINOv2, OpenAI CLIP) achieving 99.3% teacher accuracy with 297x parameter reduction and 96.4% student accuracy for real-time mobile deployment.

Foundation ModelsKnowledge DistillationPyTorchMedical ImagingDeep Learning

Brain Tumor Segmentation (BRATS 2020)

Attention U-Net architecture for precise multi-class tumor segmentation achieving Dice score 0.993. Applied CLAHE preprocessing and custom loss functions for clinical deployment.

Medical ImagingU-NetDeep LearningPyTorchImage Segmentation

AI-Powered Heart Disease Prediction System

Comprehensive cardiac risk assessment combining ensemble ML (Random Forest, XGBoost) with OCR technology achieving 95%+ accuracy. Built production Streamlit app with real-time prediction and automated PDF reports.

Machine LearningStreamlitOCREnsemble MethodsHealthcare AI

Red Light Violation Detection System

Automated traffic violation detection using YOLOv12 + ByteTrack for multi-object tracking with HSV color space analysis and geometric line-crossing algorithms for robust traffic light state classification.

YOLOv12ByteTrackComputer VisionOpenCVReal-time Detection

AI-Powered Construction Safety Monitoring

Real-time YOLOv8 web application for PPE violation detection (10 categories) with bilingual audio feedback. Full-stack Flask application with automated compliance system.

YOLOv8FlaskComputer VisionSafety MonitoringAudio Alerts

Real-Time Weapon Detection System

AI-powered surveillance system using YOLO/SSD-MobileNet for firearm and knife detection in real-time video with end-to-end pipeline for CCTV deployment.

YOLOSSD-MobileNetSecurityReal-time DetectionComputer Vision

AutoClaimVision: Vehicle Damage Assessment

CNN-based model using ResNet/EfficientNet for automated vehicle damage classification achieving 88% accuracy. Deployed via Flask API with real-time severity assessment.

ResNetEfficientNetFlaskInsurance TechDeep Learning

Real-Time Hand Gesture Recognition System

Dual-mode system combining MediaPipe Hand Landmark Detection with 7 gesture classifications achieving 95%+ accuracy at 30+ FPS with multi-hand support.

MediaPipeComputer VisionReal-time ProcessingGesture RecognitionOpenCV

Real-Time Drowsiness Detection System

Computer vision system using dlib's 68-point facial landmark predictor and Eye Aspect Ratio (EAR) algorithm with multi-stage pipeline for driver safety monitoring.

dlibComputer VisionDriver SafetyReal-time DetectionOpenCV

OpenCV Invisibility Cloak

Real-time augmented reality application using color detection and segmentation in HSV color space with dynamic thresholding for robust performance under varying illumination.

OpenCVAugmented RealityImage ProcessingHSV Color SpaceReal-time

AutoStatAgent: Multi-Agent EDA Platform

Multi-agent system with 7 specialized agents for automated exploratory data analysis. Integrated Groq LLM API with comprehensive statistical analysis and LaTeX-based PDF reports.

LLMMulti-AgentData ScienceStreamlitStatistical Analysis

Customer Churn Prediction

Ensemble ML models (Logistic Regression, Random Forest, XGBoost) achieving 92% accuracy and 0.85 ROC-AUC. Deployed via Flask API for real-time churn prediction.

Machine LearningEnsemble MethodsFlaskBusiness AnalyticsXGBoost