Avatar

Varshil Gandhi

Data & Applied Scientist

Microsoft India Pvt. Ltd.

Biography

I work as a Data & Applied Scientist at Microsoft Ads team (AI + R division), where I work on deep learning models that map relevant ads to a query. Models deployed by me are serving ads in many international markets across the world. My research interests span a variety of domains, from robotics & reinforcement learning to speech and text.

In my undergrad days at IIT Guwahati, I worked on problems at the intersection of deep reinforcement learning, imitation learning, and robotics. Both my research internships were focused on the same, where I worked under Prof. Balaraman Ravindran at RISE Labs, IIT Madras and with Dr. Venkat Natarajan at Intel Labs, Bengaluru. Apart from this, my bachelor thesis was in the speech domain, where I worked with Prof Rohit Sinha on developing universal self-supervised representations for multilingual speech systems.

Aim for the stars and lie with the soil - summarises my free time I spend with my telescope and in gardening. Please do ping me if you got some excellent astrophotography plans or gardening suggestions.

Interests

  • Reinforcement Learning
  • Robotics
  • NLP

Education

  • Bachelor's in Technology, Electronics & Electrical Engineering, 2016-2020

    IIT Guwahati

  • Minors in Computer Science, 2016-2020

    IIT Guwahati

Experience

 
 
 
 
 

Data & Applied Scientist

Microsoft India Pvt. Ltd

Jul 2020 – Present Bengaluru
 
 
 
 
 

Research Intern

RISE lab, IIT Madras

May 2019 – Jul 2019 Chennai
Under Prof. Balaraman Ravindran

  • Strived upon designing robust Imitation learning models for autonomous driving scenarios.
  • Examined how introducing graph-based structural biases can help neural networks reason about the relations between the objects present in the view of a self-driving car.
 
 
 
 
 

Technical Project Manager

Robotics Club

May 2018 – Apr 2019 IIT Guwahati
From managing the ongoing projects to upcoming projects, my responsibility was to maintain the work-flow at the club. Along with all these,

  • Project Head for ARLE team, a project under 4i Labs.
  • Designed certain problem statements and conducted those events at an institute level.
 
 
 
 
 

Research Intern

Intel Research Labs

May 2018 – Jul 2019 Bangalore
Under Dr.Venkat Natarajan, Senior Staff Engineer

  • Explored the domain of Reinforcement learning & Imitation Learning for plausible innovations towards intelligent learning of robotic systems as collaborative multi-agents.
  • Designed algorithms that can learn robotic arm maneuvers using human-demonstrated trajectories.

Recent Posts

Placement Experience | Microsoft Data Science

My on-campus placement experience and advice for those preparing for Data Scientist roles.

Projects

*

AnimalAI - Imitation Learning

Implemented a reformed architecture that uses Proximal Policy Optimization for training the agent and uses Behavior Cloning for …

ARLE- Autonomous Robot for Library Enhancement

An effort to make an automated robot capable of mananging (picking and placing) books in the institute library.

BERT on NarrativeQA

A transformer based Language model BERT trained on NarrativeQA dataset.

Fake News Detector

A fake-news detector based on BERT Language model.

Policy Gradient Algorithms

Policy gradient algorithms on Lunar lander environment

Robotic Palm Control

Robotic Palm control using hand-pose estimation algorithms on live video feed

Linear Inverse RL Algorithms

Implementation of LIRL algorithms on Mountain Car Environment.

Device for Ship Detection

A working prototype of a low cost safety device for fishing vessels to prevent collisions.

Achievements

3rd in Safety devices for Fishing vessels problem statement.

Research Fellowship

All India Rank 1915

National Level Winner

Contact