Jayant R. Kalagnanam  Jayant R. Kalagnanam photo       

contact information

Distinguished Researcher, Math Sciences,and Chief Scientist, Industrial Products, Thomas J. Watson Research Center, Yorktown Heights, NY USA
Member, IBM Industry Academy,
  +1dash914dash945dash1039

links

Professional Associations

Professional Associations:  Association for the Advancement of Artificial Intelligence (AAAI)  |  INFORMS

profile


 

Experience:

 

Chief Scientist, Industrial Products & Distinguished Research Staff Member,  March 2013- present, Mathematical Sciences, T. J. Watson Research Center, Yorktown Hts, NY 10598

Senior Manager, Aug 2007 - Feb 2013: Production Modeling, Business Analytics and Mathematical Sciences, T. J. Watson Research Center, Yorktown Hts, NY 10598

Founding Research Director, Sept 2010 - June 2012: IBM Research Collaboratory, Singapore, Focus: Analytics and Optimization for Smarter Cities.

Manager, Business Analytics and Optimization, June 2002 - July 2007, T. J. Watson Research Center, Yorktown Hts, NY 10598.

Research Staff Member Jan 1996 - current, T. J. Watson Research Center, Yorktown Hts, NY 10598.

 

Education:

Ph.D Carnegie Mellon University, Dept of Engineering & Public Policy

B.Tech. Indian Institute of Technology, Dept of Mechanical Eng

 

Areas of Expertise/Interest

AI based Solutions for Industry 4.0 & Internet of Things (IoT):

IoT enables an unprecedent real-time access to data from distributed assets and the ability to trigger actions based on inferred condition of the asset or the overall (eco)-system.  The two fundamental challenges are the ability to (i) contextualize the data streams into a meaningful semantic model, and (ii) analyze the data to learn interesting patterns that merit specific actions. 

My current focus is to build solutions for heavy industries (heavy equipment manufacturers, production plants, connected devices etc) leveraging artificial intelligence, machine learning and knowledge representation.  The first part of this effort is to understand, learn and build domain specific semantics that provide the lens needed to organize and contextualize the data streams.  This effort borrows from AI (text analysis, knowledge engineering and representation) to (automatically) build domain semantic models.  The next aspect is the use of machine learning techniques to learn from historical data (at the asset or system level) to understand the behavior and orchestration of assets - both in terms of normal and anomalous behavior.  And finally learn from data what actions are effective under which conditions to iprove asset and system behavior and outcomes.  Challenges addressed are big data analytics and machine learning targeted for sensor data.

Smarter Cities:

Science Vision - A real time, data integrated, model-based approach for the smart management of the infrastructure and resources of a city that improves resource utilization and the quality of urban services for its citizens with a focus on Smarter Cities and issues relating to environment, water, energy and traffic systems.

Condition based Asset Management (CbAM)

There are 2 components for CBAM:

  • Use of Static Data (log data)to create failure models, survival models and association rules - these are then used to create an optimal maintenance plan. This plan generate work orders that go into a work order management (system such as MAXIMO). On a weekly basis, the work orders are used to create a schedule for crew based on the location and priority of the work orders. This is the normal flow of things
  • use of Data in Motion to detect anomalies detection (near real time) monitoring data to do anomaly detection - and generate work orders (for inspection or repair) that are loaded into work order management system (e.g. MAXIMO) with appropriate priorities.
  • Analytics-driven asset management
  • Learning dynamic temporal graphs for oil-production equipment monitoring system

Smart Grids

With the advent of the Smart Grid, the infrastructure for energy supply generation and transmission is experiencing a transition from the current centralized system to a decentralized one. The ability to access real-time information on supply availability and prices supported by the demand offers unique opportunities to improve the overall efficiency of the grid in terms of both long-term supply-demand management as well as near-term dispatching of diverse generation facilities to meet current demand. The responsiveness and flexibility envisioned for the Smart Grid provides additional advantages in facing the significant new challenges of integrating distributed and intermittent generation capability such as small-scale generators and renewable energy sources (wind, solar, etc.) at a scale unmatched by current grid technology. This is becoming more critical as renewable energy technologies are playing an increasingly important role in the portfolio mix of electricity generation.

Production Modeling

Develop leading-edge solutions using combinatorial optimization problems for steel and semiconductor manufacturers. Our solutions are based on advanced techniques from operations research, and artificial intelligence combined with industry experiences.

Production Design and Operations Scheduling (PDOS) is library of optimization based algorithms for scheduling the daily operations of a steel plant. It composed of six optimization modules: Inventory Application, Slab Design, Plate Design, Melting Shop Scheduling, Hot Mill Scheduling, and Finishing Line Scheduling. PDOS has been successfully deployed at several leading steel plants around the world.