Oct 2021 – Feb 2024 · 2 yrs 4 mos · London, United Kingdom

Jan 2019 – Oct 2021 · 2 yrs 9 mos · London, United Kingdom

Jun 2017 – Aug 2017 · 2 mos · San Francisco, CA

• Tweet toxicity: Trained a recurrent neural network model that predicts how toxic a tweet is as part of a hack week project. Model was deployed into production at Twitter.
• Multi-task timelines model: Created a model that jointly predicts nine user engagements on Twitter timelines as opposed to nine separate models that existed earlier to decrease CPU load while retaining performance.
• URL embeddings: Demonstrated a method to generate embeddings of URLS using SVD on a matrix of users and URLs indicating whether a certain user visited a certain URL.

May 2016 – Jul 2016 · 2 mos · Montreal, Canada

• Supervised by Prof. Yoshua Bengio
• Implemented and compared Stochastic Gradient MCMC algorithms

Jul 2015 – May 2016 · 10 mos

• ACES-ACM is the cultural society of the computer science department of IIT Delhi as well as the official ACM student chapter of IIT Delhi.

May 2015 – Aug 2015 · 3 mos

• Summer Undergraduate Research Award (SURA) Project
• Modelling student learning using Deep Learning Techniques
• Using recurrent neural networks (LSTMs), created a model which tries to capture student learning and predict whether a student will answer the next question correctly. Implemented in Theano.