I am a first principles driven AI Research Engineer with extensive experience in designing and building  AI-powered solutions across Healthcare, Retail, Transportation and Financial Services. My research interests broadly lie in Computer Vision – particularly, the application of Deep Learning for Object Detection, Visual Recognition, and Automated Decision-Making in dynamic environments. I am particularly passionate about advancing model interpretability and robustness, with a focus on optimizing these technologies for scalable deployment. 

My recent work explores Explainable AI (XAI) methodologies for high-stakes decision-making, bridging deep learning with model interpretability through Local Interpretable Model-Agnostic Explanations (LIME), SHapley Additive exPlanations (SHAP), and counterfactual reasoning to enhance transparency and trust in AI-driven systems. I have developed real-time AI-driven computer vision systems, leveraging YOLO-based architecture for object detection and semantic segmentation, to solve high-impact challenges in Retail and Healthcare.

• In Retail, I designed an autonomous shelf-monitoring system for retail inventory optimization, enabling proactive stock replenishment through real-time product recognition. To enhance trust in AI decision-making, I implemented LIME-based heatmaps, enabling retailers to visualize which product regions contribute to stockout alerts, ensuring AI-driven recommendations are interpretable and reliable.
• In Healthcare, I developed a medical imaging analysis system that applies radiomic feature extraction across X-Ray, CT, and MRI scans to detect early implant complications in joint replacement patients. By integrating LIME-generated saliency maps, the system highlights the most critical regions influencing model predictions, offering clinicians explainable insights to guide proactive surgical intervention. This research advances AI-driven diagnostics, improving clinical decision-making, patient outcomes, and early disease detection while addressing the “black-box” nature of deep learning in medical applications.

By incorporating post-hoc explainability, I ensure that AI-driven decisions are transparent, trustworthy, and actionable, allowing domain experts to validate model predictions effectively.

As CTO of Leap2X, I collaborated with Vizuara Labs at Massachusetts Institute of Technology on pioneering multi-modal AI research, integrating vision and large language models such as CLIP (Contrastive Language-Image Pretraining) and BLIP-2 (Bootstrapped Language-Image Pretraining) to develop intelligent systems that seamlessly bridge visual perception with natural language reasoning, redefining human-machine interaction.

I have extensive experience with TensorFlow, PyTorch, OpenCV, and cloud-based AI deployment on AWS SageMaker and Google Vertex AI. My work also incorporates Scikit-Learn for traditional ML, YOLO for real-time object detection, and LIME & SHAP for explainability, ensuring interpretability in AI-driven decision-making.

I am actively involved in AI mentorship and open-source initiatives, working to bridge the gap between academia and real-world AI adoption.

Beyond research and engineering, I am an avid explorer and lifelong learner. I have traveled to 70 countries across all seven continents, immersing myself in diverse cultures and histories.

As a history buff, I enjoy engaging in socio-political conversations and deep philosophical discussions. I have been practicing Siddha Yoga meditation for over 20 years, blending mindfulness with my scientific pursuits.

When I’m not working on AI, I’m embracing the outdoors. I’ve hiked five of the world’s seven tallest peaks and hold a Nidan (Second-Degree Black Belt) in Shotokan Karate. I was the Gold Medalist at the UK National Level Shotokan Karate Championships, held at the Osaka Kyobashi Branch of the Japan Karate Association (JKA) 🇯🇵.

And when I find a rare moment to relax, you’ll probably find me strumming my six-string.