I am a machine learning reseacher based in London. My research interests are in graph neural networks, recommendation systems, and NLP. Presently, I am a Principal Applied Scientist at Microsoft, where I am part of the Microsoft Search, Assistance, and Intelligence (MSAI) team. My current focus is on neural recommendation systems using graph neural networks and transformer models. Prior to this, I was leading a team of machine learning scientists and engineers at Mastercard AI Garage where my team conducted applied research in financial domain and build fraud and risk management products. Previously, I was a researcher in the HealthCare Analytics Research Group at IBM TJ Watson Research Center in New York where I pursued research in machine learning and information visualization for healthcare. More recently, I worked at two startups Cure.Fit and Hike Messenger, where my endeavors are directed towards building machine learning based consumer products, particularly - recommendation and personalization. I have also dabbled in computational biology spending 2 years at Buckler Lab @ Cornell University where I worked on the problems at the intersection of machine learning and bioinformatics.
I dropped-out from a PhD program in pure mathematics after 3.5 years at Kansas State University. I studied algebraic & differential geometry (particularly Donaldson-Thomas Invariants and their wall crossing phenomenon). Some of my writings of this period are Geometry of Calabi-Yau Manifolds, Topological Sigma Models.
I graduated with a MMath (Part III of the Mathematical Tripos) from University of Cambridge in the UK. To read more about some of the quirks and idiosyncrasies of the mathematical tripos see here. While at Cambirdge, my interests were in string theory and black holes in higher dimensions. I wrote my Part III essay on AdS/CFT Correspondence & holography. My essay can be found here.
During my time at K-State, I became interested in machine learning in pursuit of more applied sciences. I took some ML courses and collaborated on research with a local ML research group. I developed a particular interest in NLP, graph theory, and geometric approaches for data analysis.
During my undergraduate studies in India, I have interned in theoretical physics research labs at Tata Institute of Fundamental Research - Mumbai, and at Jawaharlal Nehru Center for Advanced Scientific Research - Bangalore.
Outside the world of mathematics, my affinities are towards coffee, mountains, books, fitness, running, and cycling. I tweet at @januverma
Notes: Modern Developments in Graph Transfomers
Blog: Mathematical Introduction to Manifold Learning
Blog: Friend Recommendation at Hike
Blog: Linear Regression: Frequentist and Bayesian
Code: Graph Embedding Approaches to Recommendation
Code: Convolutional Neural Network for Text Classification in Keras
Graph Embedding Methods for Recommendation Systems, Data Hacks Summit, Analytics Vidhya, Bengaluru, Nov 2019
Transfer Learning in NLP, PyData Delhi 2019, August 2019
Beyond QWERTY: Solving the Input Problem of India, Future of Work, YourStory, Bengaluru, Feb 2019
Nuts and Bolts of Image Recognition, Software Technology Park of India, Mohali, Chandigarh, Feb 2019 Slides
ImageNet and Image Recognition, Innvovicon, Bharti Vidya Peeth, New Delhi, Feb 2019 Slides
Case for Open Data in Machine Learning, Panjab University, Chandigarh, Sept 2018 Slides
Social Bias in Machine Learning, PyData Delhi 2018, August 2018 Slides
Word Embeddings and NLP, Netaji Subhash Insititute of Technology, April 2018 Slides
Data Driven Healthcare, Indraprastha Institute of Information Technology (IIIT), New Delhi, Dec 2017 Slides
Improving Healthcare with Data Science, Facebook Dev Circle, New Delhi, Oct 2017 Slides
Data Science in Healthcare, Cluster Innovation Center, University of Delhi, Sept 2017 Slides
Understanding Clustering: Supervising the Unsupervised, PyData Delhi 2017, Sept 2017 Slides
Word Vector Representations (word2vec) and Text Classification, ML-India, Jan 2017, Slides
Case for Data Visualization, DataViz Delhi, Nov 2016 Slides