Spectrum of Portfolios

Source Code

Click for Image description

• I am an Founding Machine Learning Engineer at Veer, an Oxford spin-off founded by renowned AI researcher Sir Nigel Shadbolt and former newspaper executive Sue Douglas. Previously, I was a Machine Learning MSc graduate student and Google DeepMind Scholar at University College London, where I specialized in Bayesian deep learning, NLP (alignment and evaluations), Multi-agent AI, Mechanistic interpretability, Kernel methods, and Algorithmic trading. Before this, I worked as a Deputy Manager and Machine Learning Architect at Ford Motor Company, where I led a small team to develop evaluation frameworks for ML models. These evaluations measured computational cost, memory consumption, and time efficiency for ML models built for Ford Credit worldwide. I also focused on assessing the safety and ethical considerations in the models' decision-making processes. I also served as a Quantitative Researcher at Trexquant Investments LLP (AUM $2B), applying quantitative and ML methods to build optimized equity portfolios in liquid markets. In addition, I interned at Elan Capital Management LLP focusing on modeling - yield curve dynamics and HFT trades using machine learning models, and building an intelligent parser leveraging LLMs.


• I am the recipient of the prestigious Google DeepMind Fellowship, the Endeavour Research Fellowship awarded by the Australian government, and CRISIL's Best Doctoral Research Award in Investment Management (AI in Finance-Track). I hold a Ph.D. in Computational Finance from IIT Madras, supervised by Prof. M. Thenmozhi (IIT Madras). My thesis chapters have been published in journals like European Financial Management, Finance Research Letters, Research in International Business and Finance, Carbon Management, and Decision. My dissertation focused on building statistical and ML models aimed at addressing the longstanding enigma in finance—how to explain merger waves? I collaborated with Dr. Chinmoy Ghosh (University of Connecticut), Dr. May Hu (RMIT University), Dr. Robert Faff (University of Queensland), Dr. Syed Shams (University of Southern Queensland), and Dr. Veerasamy Sejian (University of Queensland).


• My expertise spans a wide range of machine learning tools, including Python libraries such as Pandas, NumPy, PyTorch, TensorFlow, Pyfolio, Riskfolio, PyMC3, OpenAI Gym, Scikit-survival, and Lifelines. I am proficient in fine-tuning LLMs with Hugging Face Transformers, optimizing GPU workloads using CUDA, and scaling AI models through distributed training with Ray, Horovod, and DeepSpeed. I have strong skills in hyperparameter optimization using Hyperopt and experience with containerized deployments through Docker and Kubernetes on Google Cloud Platform (GCP). My proficiency with Git ensures efficient version control and seamless collaboration in large-scale projects.


• Based in London since 2023, I spend my spare time writing a blog focused on Deeplearning, Multiagent and AI alignment titled ‘myAInotes’. I enjoy trekking and volunteer to teach math and computing to tribal girl children through Tinklesevaa classroom sessions. I also promote diversity and inclusion through my sketches, which you can explore in my Behance profile or gallery section.