CV
Education
- MSc in Computer Vision, MBZUAI, Abu-Dhabi, 2024–2026
- Supervised by Ivan Laptev
- GPA: 3.75/4.0
- Coursework: Human and Computer Vision, Geometry for Computer Vision, Visual Object Recognition and Detection
- Thesis topic: Human Object Interaction Synthesis
- BEng in Electrical Engineering, SEECS, NUST
- Supervised by Dr. Wajahat Hussain
- GPA: 3.45/4.0
- Coursework: Computer Vision, Machine Learning, Digital Signal Processing
Work experience
Computer Vision Intern
VisionLabs, Remote Jun 2024 - Jul 2024
- Implemented LoRA-X to enhance Tiny-Vit models, enabling the learning of new attacks while preserving knowledge from previous models.
- Trained deepfake detection models on Veo3 and KlingAI datasets.
Data Scientist
Softtech Solutions, Islamabad, Pakistan
Jan 2024 – Aug 2024
- Designed, trained, and optimized a Document AI pipeline using PyTorch and TorchServe to extract structured data from wine menus with high accuracy
- Managed and enhanced a comprehensive database of over 500,000 wine entries, ensuring data integrity and scalability
- Developed and deployed a robust backend system using Django, enabling seamless integration with front-end applications and efficient data processing
- Spearheaded the deployment process, transitioning from AWS EC2 to Hyve Solutions servers, improving system performance and reducing costs
Machine Learning Engineer
Remote
Jan 2023 – Aug 2024
- Engineered a Deep Learning model capable of extracting structured wine information from diverse PDF templates, achieving reliable performance across over 10,000 data points
- Deployed a Docker Compose application integrating Django, MySQL, and TorchServe APIs to an AWS EC2 instance
- Built a stock price forecasting model in TensorFlow, delivering high accuracy and actionable insights for financial predictions
- Used Amazon Textract and PyTorch to digitize and analyze pre-1900 documents, preserving historical data with machine learning
Machine Learning Engineer
DCube Tech, Islamabad, Pakistan
Jun 2022 – Jan 2023
- Updated and optimized multimodal PyTorch models for Document AI
- Built and maintained data and model pipelines using DVC
- Enhanced data processing systems by analyzing and debugging 5000+ lines of code
- Helped deploy models that served over 800 clients
Data Analyst
UniVision, Islamabad, Pakistan
Jun 2022 – Nov 2022
- Collected, cleaned, and analyzed undergraduate admission data for over 500 universities globally
- Researched and identified valuable sources of university data
- Performed quality assurance for over 100 university profiles
- Led a small data mining team of two analysts
Research Intern
ROMI Labs, Islamabad, Pakistan
Jun 2022 – Nov 2022
- Contributed to projects involving machine learning, data encryption, and adversarial attacks in virtual environments
- Executed JSMA attacks on ML models and implemented AES encryption within Minecraft
- Worked with Microsoft’s Malmo project to create user-specific virtual worlds
Projects
Exploring Blindness of MLLMs (AI701)
- Conducted research to identify limitations in the visual capabilities of state-of-the-art multimodal language models (MLLMs)
- Tools: Python, PyTorch, HuggingFace
Medical Document Information Extractor
- Developed a BERT-based deep learning model for named entity recognition in medical documents
- Tools: Python, PyTorch, DVC, Git, AWS Textract, HuggingFace
Bachelor’s Thesis: Robot Policy using Machine Learning (2022)
- Built a hand gesture recognition system using the HANDS dataset and OpenCV
- Enabled a robot vehicle to receive commands via Bluetooth based on hand gesture inputs
- Tools: Python, OpenCV, Arduino
Skills
- Machine Learning Stack: PyTorch Lightning, Hydra, PyTorch, Scikit-learn, TensorBoard, TorchServe, TFX, HuggingFace
- Python Stack: NumPy, Pandas, Matplotlib, Selenium, OpenCV, TensorBoard, Flask, Django
- Database Stack: MySQL, PostgreSQL, DBeaver
- Version Control: Git, DVC
- Misc: GCP, AWS
