About
Email: atharmas [AT] purdue [DOT] edu
I am currently looking for full-time opportunities.
I am Aathman Tharmasanthiran, a Master's student in Computer Science at Purdue University with 6+ years of combined industry and research experience in AI/ML. Currently, I am working as an AI Software Developer at Lyohub, where I built a multimodal RAG chatbot that retrieves and responds with relevant images, video clips, and text from diverse knowledge sources on pharmaceutical manufacturing. This system is fully deployed in a Kubernetes environment to ensure scalability.
Previously, I worked as a Software Engineer - Machine Learning at WSO2 for over 2.5 years, where I led the end-to-end development of multiple production-grade applications— gathering requirements, building, testing, and deploying with CI/CD in Kubernetes environments. I have significant experience working in agile teams following Scrum methodologies.
Overcoming challenges: Since Vision Pro development lacks proper tutorials and documentation, I had to navigate uncharted territory—manually parsing technical docs, experimenting with APIs, and solving issues that even modern LLMs couldn’t assist with. This project demonstrates my ability to self-learn, adapt, and innovate when facing cutting-edge, under-documented technologies.
My interests span across LLM applications, Retrieval-Augmented Generation, Generative AI, Robotics and Multi-Agent Systems, Computer Vision, Natural Language Processing, and Reinforcement Learning.
My current research focuses on developing a Vision-Language Action model that learns from human demonstration videos rather than traditional teleoperated imitation learning. This approach reduces data collection costs and leverages abundant internet video data to enable more scalable robot training.
I received my Honors Degree of Bachelor of Science Honors in Engineering (B.Sc.(Hons.) in Engineering) specializing in Computer Science and Engineering from University of Moratuwa, where I completed my thesis on Multi-Aspect Deep Learning approach for Human Affect and Behaviour based Threat Prediction and Video surveillance under the supervision of Prof. Damminda Alakahoon, Dr. Daswin de Silva, Dr. Rashmika Nawaratne and Mrs.Vishaka Nanayakkara.
During my undergrad, I did an internship at WSO2 LLC, Sri Lanka (Branch of WSO2 headquarter located at Santa Clara, CA, USA) as a Research Intern working on the automatic tuning of thread pool sizes of Ballerina microservices by taking into account the real-time load configurations and feeding them into a Gaussian Process Bayesian optimization model improving the 99 percentile latencies of those microservices in realtime. This research had the potential of improving the performance of microservices that are deployed in production giving a drastic improvement in the overall performance of the system approximately upto 26%.
Upon graduation, before starting my graduate studies, I worked as a Software Engineer at WSO2 LLC, Sri Lanka (Branch of WSO2 headquarter located at Santa Clara, CA, USA) where I researched, developed and maintained multiple AI applications for a product of WSO2 called Choreo.
News!
- November 2024: Joined LyoHub at Purdue University as a Graduate Student Researcher
- May 2024: Joined e-Lab at Purdue University as a Graduate Student Researcher
- September 2023: Joined CoRAL Lab (Cognitive Robot Autonomy & Learning Lab) at Purdue University as a Graduate Student Researcher
- August 2023: Started my Masters in Computer Science at Purdue University.
- April 2021: Started my first job as Software Engineer-Machine Learning at WSO2 LLC, Sri Lanka (Branch of WSO2 headquarter located at Santa Clara, CA, USA)
- September 2021: Our paper Minority Resampling Boosted Unsupervised Learning with Hyperdimensional Computing for Threat Detection at the Edge of Internet of Things to IEEE Access vol. 9
- July 2021: Graduated from the Faculty of Engineering, University of Moratuwa, Sri Lanka with a first class in BSc.(Hons.) in Computer Science and Engineering.