At MPOnline Limited, I architected diagnostic pipelines for brain tumor MRI datasets using custom CNNs and MobileNetV2 transfer learning, achieving over 90% classification accuracy.
I also engineered a Retrieval-Augmented Generation assistant with LangChain and ChromaDB, and implemented DQN, Dueling DQN, and PPO architectures for continuous control benchmarks. For production workflows, I containerized inference microservices and built asynchronous REST APIs with CI/CD deployment pipelines.
On my AI-Driven Healthcare Web App project, I combined disease-prediction classifiers with a conversational interface that extracts symptoms from free-text input. I’m also building PickleballVision, a computer vision tool that classifies shots from video using pose estimation and joint-angle analysis.

