United Kingdom
Reimagining healthcare with an AI-native operating system. At Enthara, we were building, the clinical operating system that gives every clinician a live, unified, intelligent view of their patient and a set of agentic tools that automate the work no one has time for. My work span across: • Shaping the vision, roadmap, and product strategy • Working closely with clinicians to design human-centered workflows • Architecting our shared patient context layer and agent platform • Building a best in class team across engineering, design, and research • Leading fundraising, partnerships, and GTM
Working on Machine Unlearning in generative models. Built GeneticBPE - A novel tokenizers for miRNA sequences. Papers accepted at TokShop ICML'25, MUGen ICML'25
Conducting my Master’s thesis jointly at UCL and The Alan Turing Institute under Dr. Tapabrata Chakraborty and Dr. Gary Royle. Developing a multimodal FMs for dermatology. Implementing test-time adaptation techniques to improve model robustness on out-of-distribution. -> Work accepted at CVPR ViSCALE'25. Designing and constructing dermatology knowledge graphs, combining open-source ontologies and image annotations to represent lesion taxonomy and relationships Curating and annotating a Visual Question Answering (VQA) dataset for dermatology from public sources, enabling image-driven clinical Q&A workflows
-> Improved spectral resolution in Ultrasound images by combining beltrami geometry with Tformers. -> Designed dermatology knowledge graphs, combining open-source ontologies and image annotations to represent lesion taxonomy and relationships. -> Implemented test-time adaptation techniques to improve model robustness on out-of-distribution. -> Developed KG enabled MoE for Dermatology with skin color grading.
Collaborated with Prof. Dr. Arun Kumar Dubey in the Soft Computing Group to advance AI-driven healthcare solutions. Developed COVLIAS 4.0, a novel feature-fusion and weighted-average ensemble framework that improved COVID-19 diagnostic accuracy by 8%. Engineered an attention-modulated text-analysis pipeline for Depression detection, achieving 90% precision in identifying depressive indicators in patient narratives. Designed and trained deep-learning architectures (CNNs, RNNs, Transformers) for medical imaging and clinical-text applications, optimizing performance through hyperparameter tuning and data-augmentation strategies. Authored detailed experimental reports, presented findings at weekly lab seminars, and contributed to manuscript preparation for peer-reviewed publication.