Sep 2020 – Aug 2023 · 2 yrs 11 mos · New York, New York, United States

• Led projects across the firm for time-series, natural language processing, LLMs and generative AI.
• Time-series signal processing with Neural ODEs, Temporal Fusion Transformers for 1) real-time bond pricing - built the current SOTA product in the market 2) pricing mortgage-backed securities 3) hedging commodities swaps 4) forecasting oil pipeline flows 5) anomaly detection in debit transactions
• NLP projects: 1) LLMs for predicting market moving signals from news 2) Anomaly detection in debit transactions 3) Analyzing data across trading chat rooms in real-time for quantitative insights

Oct 2018 – Aug 2020 · 1 yr 10 mos · Mountain View, California, United States

• Led applied machine learning projects at Google X (the moonshot factory). Developed AI applications in domains including sensor fusion and signal denoising, 3D scanning/reconstruction (patent), genetic engineering, audio/speech processing, motor control optimization for robotic knees, EEG decoding, and forecasting energy consumption in the power grid

May 2018 – Sep 2018 · 4 mos · Mountain View, California, United States

• Worked at Google X, Alphabet's moonshot factory, on applied machine learning and AI research as part of the central machine learning team. Developed an ML solution for improving the wind, pressure, and temperature sensors in extreme atmospheric conditions for Project Loon (patent) and contributed to two other early-stage research projects

Jun 2017 – May 2018 · 11 mos · Champaign, Illinois, United States

• Developed a deep learning model based on sequence-to-sequence LSTMs with attention layers for converting natural language queries to code. Trained this model on billions of previous natural queries and improved the WolframAlpha computational knowledge engine

May 2016 – Aug 2016 · 3 mos · Los Alamos, New Mexico, United States

• Selected as a Co-Design Fellow. Worked on HPC Smoothed Particle Hydrodynamics (SPH) simulations of neutron star mergers for studying kilonovae and r-process nucleosynthesis in the ejecta using supercomputers.
• Augmented the 2HOT HPC library by adding tabulated equations-of-state and optimized the nearest neighbor search algorithm by implementing domain-partitioning schemes based on Hilbert ordering using Charm++ and STAPL

Aug 2015 – May 2018 · 2 yrs 9 mos · University of Illinois at Urbana-Champaign

• Pioneered the application of deep learning in gravitational wave astrophysics as a member of the Nobel-prize winning LIGO group. Developed the first deep convolutional neural networks trained with gravitational waveforms and numerical relativity HPC simulations of merging black holes using the Einstein Toolkit on the petascale Blue Waters supercomputer, to build a real-time gravitational wave detection pipeline for LIGO.
• Published 5 first-author peer-reviewed papers and over 20 co-authored articles. Won the ACM Student Research Competition, the NVIDIA Graduate Fellowship, the CSE Fellowship, as well as several best paper and poster awards

Jul 2014 – Jun 2015 · 11 mos · IIT Bombay, Mumbai, India

• Worked on computational nonlinear dynamics and chaos research projects with Prof. Punit Parmananda. Wrote senior undergraduate thesis on numerical simulations of Chimera States in systems of coupled mercury beating heart oscillators.

May 2014 – Jul 2014 · 2 mos · Notre Dame, Indiana, United States

• Worked on numerically analyzing the magnetohydrodynamic stability of astrophysical jets with Prof. Dinshaw Balsara and his group. Performed both symbolic computations with Mathematica and numerical simulations on parallel HPC clusters in C++/Fortran. Co-authored a journal publication