Research Interests

Deep Reinforcement Learning, LifeLong Learning, Deep Learning, Big Data Analytics

Research Experience

On Training Recurrent Neural Networks for Lifelong Learning

2nd Continual learning Workshop, NeurIPS 2018
Shagun Sodhani*, Sarath Chandar*, Yoshua Bengio

Environments for Lifelong Reinforcement Learning

2nd Continual learning Workshop, NeurIPS 2018
Khimya Khetarpal*, Shagun Sodhani*, Sarath Chandar, Doina Precup

Compositional Language Understanding with Text-based Relational Reasoning

Relational Representation Learning Workshop, NeurIPS 2018
Koustuv Sinha, Shagun Sodhani*, William L Hamilton, Joelle Pineau
Relational Representation Learning Workshop, NeurIPS 2018

On Capacity Expansion in Recurrent Neural Networks for Lifelong Learning

Lifelong Learning Workshop, ICML 2018
Shagun Sodhani*, Sarath Chandar*
Sanket Vaibhav Mehta*, Shagun Sodhani*, Dhaval Patel
Shagun Sodhani*, Vardaan Pahuja*

Industry Experience

Member of Technical Staff II

June 2016 - September 2017

  • Teaching Assistant for the Machine Learning course for employees at Adobe Systems.
  • Won the Outstanding Young Engineers Award at Adobe Systems. The award recognises young engineers displaying outstanding technical acumen and innovation.
  • Worked to automate the process of converting scanned forms into adaptive forms by understanding the semantics of the form.
  • Co-mentored two summer internship projects (2017) related to Structure Extraction from scanned documents and Visual Question Answering.
  • Prototyped a conversational agent to automate tech support conversations.
  • Co-mentored an internship project (2016) to model Reinforcement Learning based Conversational Agents for usecases like browsing an ecommerce site. This work was presented at Adobe Tech Summit (2017).
  • Member of Technical Staff I

    June 2015 - June 2016

  • Developed application for mining interesting insights from multi-channel data. This work was presented at Adobe Data Scientist Summit (2016) and Adobe Tech Summit (2017).
  • Developed an interactive data exploration tool on top of Spark to assist data scientists on Adobe’s Data Platform.
  • Developed a web service layer over Analytics Premium (DataWorkBench) with optimisations like sharding, caching sub-query results, and streaming partially aggregated results to support thin clients like web dashboards.
  • Research Intern

    May, 2014 - July 2014

  • Worked in the area of Instore User Experience Enhancement to improve foot-fall in brick and mortar stores.
  • Used the store’s context to model and propose solutions based on Internet of Things, and Bluetooth Low Energy devices.
  • Teaching Experience

    Data Mining Techniques, HEC, Montreal

    September, 2018 - December, 2018
    Teaching Assistant

  • Conducted the programming sessions about machine learning and data mining tools.
  • Created assignments and conducted office hours.
  • AI for Social Good Lab

    May 2018 - June 2018
    Mentor

  • Mentored students in machine learning, prototype development, and design skills.
  • Machine Learning Program for Adobe Employees

    2017
    Teaching Assistant

  • Worked as a teaching assistant for a Machine Learning course that was offered to employees at Adobe Systems
  • Databricks, edX

    June, 2016 - September, 2016
    Teaching Assistant

  • Worked as a teaching assistant for the 3-course series focusing on Data Science and Engineering with Spark XSeries.
  • The course was created in partnership with professors from University of California, Berkeley, University of California, Los Angeles and Databricks and was offered on the edX platform.
  • Helped in creating assignments for the students and assisted them on the Piazza.
  • Projects

    A-Paper-A-Week - Committed myself to reading and summarising a research paper every week.
    Character-Level Language Model - TensorFlow based implementation of Char-RNN using Multi-layer Recurrent Neural Networks.
    Spatial Co-location Pattern Mining using Graph Database - Developed and implemented scalable algorithms for mining spatial co-location patterns using distributed graph database as part of the final year thesis.
    Stack Exchange Data as Spark Dataframes - Implemented a system that loads Stack Exchange Data as Spark Dataframes and performs analysis over the data.
    Tagging StackExchange Questions using SVC - Implemented a tagger for questions posted on Stack Exchange using Support Vector Classification as course project for Artificial Neural Networks, IIT Roorkee.

    Workshop and Talk Organization

    Extracurricular Activites

    Member

    March 2017 - September 2017

  • Worked to spread awareness about open source practices & guidelines across Adobe Systems.
  • Co-initiator, Secretary and Treasurer

    January 2014 - December 2014

  • Co-initiated ACM Student Chapter as the first departmental computer science society for students at IIT Roorkee.
  • Worked to promote the knowledge of computer science by providing students with a platform to interact with distinguished academicians and professionals.
  • Secretary

    April 2013 - March 2014

  • Problem setter for different programming contests including IIT Roorkee's annual algorithmic programming contest called Insomnia.
  • Member

    January 2012 - May 2015

  • Worked on and mentored various student driven software projects.