Shubham Agarwal

Research Software Engineer - AI

MIT-IBM Watson AI Lab, Cambridge, MA


I am currently a Research Software Engineer - AI at the MIT-IBM Watson AI Lab, Cambridge, MA. My main interests lie in the intersection between AI, dialogue systems and planning. Some of the projects I have completed include creating intelligent conversational agents, declarative dialogue design and execution systems, and automated time-series analytics (Award-winning project as Best Technical Demo at AAAI 2020).

I hold a Master’s degree in Computer Science from Arizona State University in July 2017, where I worked on Natural Language Understanding and Semantic Parsing of English natural language (see thesis: “ Aligning English Sentences with Abstract Meaning Representation Graphs using Inductive Logic Programming “).


  • Artificial Intelligence
  • Dialogue Systems
  • AI Planning
  • Kubernetes, Docker


  • Masters in Computer Science, 2017

    Arizona State University, Tempe

  • Bachelors in Information Systems, 2015

    Birla Institute of Technology and Science, Pilani



Dialogue Systems

AI Planning






Research Software Engineer

MIT-IBM Watson AI Lab

Aug 2017 – Present Cambridge, MA
Dialogue Systems, AI Planning, Cloud deployments and scaling, Kubernetes, Docker

Masters, Computer Science

Arizona State University

May 2015 – Aug 2017 Tempe, AZ
Thesis in NLP | Instructor: Object Oriented Programming

Bachelors, Information Systems

Birla Institute of Technology and Science, Pilani

May 2011 – May 2015 Pilani, India




Generating Dialogue Agents via Automated Planning

Model Acquisition Interface

A declarative approach to dialogue design.


A Platform to bridge gap between State-Of-The-Art Time-Series Analytics and Datasets

Recent Publications

TraceHub - A Platform to bridge gap between State-Of-The-Art Time-Series Analytics and Datasets

(AAAI Best Technical Demo Award!) A platform that connects new non-trivial state-of-the-art time-series analytics with datasets from different domains

Bayesian inference of linear temporal logic specifications for contrastive explanations

Contrastive explanations using BayesLTL – a Bayesian probabilistic model for inferring contrastive explanations as linear temporal logic (LTL) specifications

Planning for Goal-Oriented Dialogue Systems

A declarative representation of the dialogue agent to be processed by state-of-the-art planning technology

Executing Contingent Plans: Addressing Challenges in Deploying Artificial Agents

A proposed executor that can reason using the sophisticated action effects, and we demonstrate the impact this can have empirically


  • agarwalshubham2007@gmail.com