I fell in love with programming in 6th grade when I learned that my friend could develop the games I played on my PC. I attempted to create a Visual Basic application in high school, but it was unsuccessful. My passion for programming persisted, even after graduating with a mechanical engineering degree. I adore reading, writing, thinking about, improving, and understanding programs, especially in Python.

I am proficient in TensorFlow (flexible to work with PyTorch, and Jax), and I am passionate about creating useful solutions using machine learning.

Outside of work, I enjoy reading 📚, playing badminton 🏸 and ping pong 🏓, and learning new things 🧠.

Publications

• May 2023: HolisticDFD: Spatiotemporal Transformer Embeddings Infusion for Deepfake Detection (Information Sciences)
• April 2023: Multimodaltrace: Deepfake Detection using Audiovisual Representation Learning (CVPR 2023 Workshop on Media Forensics)

Projects

• Facial Recognition Deployment with PostgreSQL and pgvector: Deployment of a FastAPI application that utilizes PostgreSQL as its database backend, with support for the pgvector extension for facial recognition. It includes functionality for extracting and saving facial embeddings to PostgreSQL, as well as comparing new embeddings to find the closest facial embeddings.
• K3-Addons: Additional multi-backend functionality for Keras 3. Multibackend attention mechanisms, advanced optimizers, or specialized layers, etc in Keras 3.
• K3IM: Keras 3 Image Models Implementation of Image Models and their 1D/3D/Space-Time extensions in Keras 3 (more than 30 in total). The models have been trained and tested with TensorFlow, PyTorch and JAX backends. (repo)
• Keras Core: Open-source contributions include implementing cross-backend patch extraction for vision transformer-like models and porting examples to Keras Core (merged pull requests).
• Deep Forgery Detector: Designed and developed Deep Forgery Detector, an extensible framework for deploying visual deepfake detection algorithms (work done at SMILES Lab @ Oakland University).
• Forensic Examiner: Designed and developed Forensic Examiner, an extensible framework for benchmarking and evaluating audio spoofing detection algorithms (work done at SMILES Lab @ Oakland University).
• FocalNet Keras 3: Multi-backend implementation of FocalNet for Keras 3. Works with TensorFlow, PyTorch, and Jax. (link).
• FocalNet TensorFlow: Implemented and ported weights for Focal Modulation Network in TensorFlow/Keras (link).
• StorySeed: Developed StorySeed, a Hugging Face Space that generates story ideas from images (HuggingFace Spaces).
• Evaluating Modern Vision Architectures on 3D Biomedical Data: Published a Colab notebook and Google Slides presentation that compare the performance of different vision architectures on 3D biomedical data (Colab Notebook, slides).
• Oil Debris Monitoring System using deep learning: Led a team of five members to fabricate an oil debris monitoring sensor that employed time-series anomaly detection techniques to report anomalous patterns in lubricants. The project has its significance in preventive maintenance for moving mechanical systems (work done at UET Taxila).

Updates

• October 15, 2022: Spoke at UET Taxila on “An Introduction to Machine Learning for Engineers.” (Slides)