top of page





After graduating from my Masters in 2010 and before rejoining school at Cambridge in 2015 I worked as a quant at Credit Suisse. My work here entailed fine-tuning their electronic market making book in FX, here I used a lot of traditional statistical modelling like kalman filters and time series. I then worked as a high-frequency trader at Citadel, a Chicago based hedge fund. I have been lucky to work with some really smart people at these firms.

I graduated from high school in 2005 and enrolled for a double major in Math and Economics at the University of London External System (now called International programmes). I self-studied for my undergraduate degree (3 years!), mainly from home and  corresponding with mentors in London until 2008.


I graduated with Distinction.

In 2009, I left India to study for a M.Sc in Applied Math at LSE. My main interest here was Cryptography and Information Science.

I graduated with Distinction.




I was slowly realizing that things outside the quant world were changing and we were just not catching up fast enough. The statistics we used to design our models was outdated and a range of emerging techniques within machine learning looked promising but very few people in the quant world understood. Besides, regulation was getting tighter and it became harder to make the case for innovation. 

A combination of discontent combined with my latent desire to go back to school (where I have felt most at home) made me apply to Cambridge! To my surprise I passed the interviews but before embarking on a Ph.D I had to prove I could still do math and write code.

I graduated with Distinction, my degree was called - MPhil in Scientific Computing.

In 2017 I started my Ph.D in the department of Physics. I am fully funded by the Alan Turing Institute which is UK's national institute for the study of data using machine learning.


Theoretical machine learning deals with inventing new variants of existing machine learning tools to make them more effective or perform better in a computational sense. I have narrowed my theoretical focus to a sub-field of machine learning - probabilistic machine learning (PML) (also called Bayesian non-parametrics). As the name suggests, the result of inference using PML is a probability distribution rather than a point estimate, in this sense it quantifies uncertainty in  predictions.

A measure of statistical guarantee or confidence in machine learning outputs is becoming an increasingly important requirement in their usage. Bayesian non-parametrics is the only current paradigm that allows the user to stipulate a posterior belief or a prediction in terms of a probability.

My supervisors are Dr. Anita Faul and Dr. Christopher Lester.

I collaborate with Prof. Richard McMahon who currently heads up the Institute of Astronomy, where we are trying to use machine learning techniques to identify gravitationally lensed quasars. 


bottom of page