Saisamarth T.

WORK HISTORY

EDUCATION

ABOUT SAISAMARTH T.

I am an AI Engineer at Danucore and a Machine Learning researcher specializing in Agentic AI, LLM Reasoning, and Distributed Inference/Training. I recently graduated with Distinction from the University of Edinburgh with an MSc in Speech and Language Processing.Current Focus: At Danucore, I work in R&D, exploring novel research ideas to enhance the performance and efficiency of LLMs. My role involves developing autonomous agentic frameworks and enhancing LLM reasoning capabilities, while also working on managing distributed training strategies on large-scale GPU clusters.Background: My research journey is driven by a curiosity to make AI systems not just powerful, but also reliable and efficient. During my Master\'s, I tackled the challenge of error propagation in complex reasoning by developing a framework that trains LLMs to intrinsically self-correct their mistakes and, crucially, to determine when they have reached the right answer. Prior to this, my research at the Indian Institute of Science approached efficiency from a hardware perspective. I co-authored papers like PowerTrain and Fulcrum that utilize transfer learning and intelligent time-slicing to optimize concurrent DNN training and inference on edge accelerators, balancing speed against power budgets.Industry Experience: Beyond academic research, I have engineered scalable AI solutions across various domains. My experience includes architecting RAG-based evaluation systems and automating high-volume financial document processing by fine-tuning models. I have also developed synthetic data generation pipelines and predictive analytics for stock market trends.My hobbies include video games, anime, and trekking. I am always open to discussing new developments in AI and collaboration opportunities. Let’s connect!