Vivek Mathur
Research Fellow at M365 Research and Microsoft Research India
About
I am a Research Fellow at Microsoft Research India, where I specialize in the scalability and trust of Large Language Model (LLM) inference systems. My primary research involves developing agentic LLM scheduling strategies to manage complex, multi-round tool-calling workflows.
Prior to joining Microsoft, I held several key positions in the industry:
-
Salesforce Einstein: Engineered real-time anomaly detection for the Einstein Trust Layer using Spark and reinforcement learning.
-
Tower Research Capital: Developed a multi-agentic LLM chat platform for code comprehension and a RAG-enhanced pipeline for quantitative trading signals.
-
IBM Research: Focused on large-scale multi-modal modeling and ESG analytics using Granite time-series models.
I hold a B.Tech in Computer Science & Engineering from IIIT Hyderabad and am currently continuing my research there as an MS (Research) candidate. My academic honors include being named a KVPY Fellow and appearing on the Dean’s List (top 5% of my batch). Beyond systems research, I am passionate about using AI for social good, as seen in my work on “UnSight,” a platform designed to bridge the educational gap for visually impaired learners.
Featured content
Unsight: Bridging The Educational Gap For Blind Learners
Visually impaired learners face significant challenges in accessing quality education due to the lack of accessible educational materials and adaptive learning environments. Traditional educational methods often fall short in addressing the unique needs of these learners, leading to a gap…
Development of sensor for monitoring the integrity of the lubrication system of heavy duty pumps in nuclear reactors
In Prototype Fast Breeder Reactor(PFBR), heavy secondary sodium rotary pumps are provided with oil pots to collect the lubricating oil of the cooling system in case of minute leaks. These leaks need to be detected as early as possible. The…