published Online Urdu Text-Line Recognition Turning pen dynamics into image channels so pretrained offline recognizers can use them. First-author paper at ICDAR 2026. JPEG-Induced Bias in Document Forensics Document-tampering detectors rely on the JPEG 8×8 grid, not the edit. Two simple attacks expose this. BMVC 2026, equal-contribution first author. ongoing Foundation-Model Distillation for Edge Inference Distilling large vision encoders into compact students that run on a Raspberry Pi, evaluated on pest detection. Final-year project, in progress. Offline-to-Online Handwriting Trajectory Recovery Recovering pen trajectories from static handwriting images to help low-resource recognition and cross-language transfer. In progress. Generative Document Forgery with RL Using reinforcement learning to steer image-generation foundation models toward realistic document forgeries for robustness testing. In progress. engineering DiffUTE Reproduction A paper-faithful reproduction of DiffUTE (NeurIPS 2023), a diffusion model for universal scene and document text editing. Adaptive Distributed SGD A runtime controller that switches between Parameter-Server and Ring-AllReduce training based on measured latency and worker failures. rh: Reproducible Research on HPC A tool for keeping code, datasets, experiments and results organized across projects, on a laptop and on a Slurm cluster. In development.
