Ashish Jagmohan  Ashish Jagmohan photo         

contact information

Principal Research Staff Member
Thomas J. Watson Research Center, Yorktown Heights, NY USA



I'm a Principal Research Staff Member at the IBM T.J. Watson Research Center in NY. My technical interests include machine learning, deep learning, reinforcement learning, Bayesian inference, natural language processing, and supply-chain optimization.

Over the last few years, I have led AI and analytics for IBM Blockchain Solutions. We created algorithms for monitoring end-to-end supply-chain performance, detecting IOT and supply-chain exception events, and predicting disruptions using Bayesian inference and classical ML models, at scale (batch-processing hundreds of GBs of data, and tens of millions of supply-chain events per hour). At the same time, I also led a forward-looking multi-year collaboration with AI and OR researchers at IBM Research and academia on using AI and stochastic optimization for multi-echelon business networks, demonstrating significant cost/revenue gains in multiple settings (perishable waste-reduction, online-groceries, on-time delivery, trace-recall). We created new techniques for decision making in multi-echelon networks, by combining deep reinforcement learning with mixed integer programming, demonstrating superior performance over both popular operational heuristics and state-of-art deep learning baselines.

Prior to this, I was a researcher and technical lead in the Business Solutions and Mathematical Sciences organization in Services Research. We created machine learning and natural language processing algorithms and systems in engagements with external clients and internal business-unit partners across several sectors: education, consumer goods, compliance, digital marketing, services, and talent and skill management. We leveraged structured and unstructured data at scale, using ML ideas from deep-learning, Bayesian inference and classical ML, and recommender systems, and natural language processing ideas including neural embeddings and concept-graphs. Our technology was deployed at multiple external clients and internal partners, and resulted in 2 Corporate Technical Awards (IBM's top technical award), E-level and O-level Accomplishments, and multiple papers and patents.

In the past, I have also worked on memory and storage systems, designing new controller and system-level techniques for enhancing reliability and performance for cutting-edge phase-change memory, NAND Flash and DRAM systems. We used ideas from information theory, algebraic and rateless coding theory, and computer architecture. This effort resulted in multiple IBM Research Division Awards and impact on multiple IBM systems product lines.

I have also worked on the use of information and learning-theoretic techniques to drive fundamentally new video compression and communication paradigms. Our research resulted in the creation of standard-compliant (H.264) and distributed source coding video codecs which were state-of-the-art in both industry and academia. This effort resulted in an IBM Outstanding Innovation Award.

I am an IBM Master Inventor, and have more than ninety granted patents and filed patent applications. I have more than sixty peer-reviewed publications in the above fields. I received my Ph.D. from the University of Illinois, Urbana-Champaign and my B.Tech. from the Indian Institute of Technology, Delhi.