- Document tagging and summarization with LLMs- Develop financial time-series model and develop explainability toolkit to interpret the model behavior- Train and deploy NLP models to extract insights from unstructured chats/emails

Deep Learning Approach for Asynchronous, Multi-variate Time Series Financial Data- Devise a new model that utilizes Neural ODE to treat temporal evolution continuously, and a modified recurrent network GRU to update the hidden states with Asynchronous incoming data.- Achieve state-of-the-art results compared to the model in production which handles one single time-series at a time.

Understand the formation and evolution of Supermassive Blackholes through Machine Learning- Develop a new research platform -QuasarNet that will provide centralized infrastructure support for innovative modelling of supermassive blackholes leveraging Machine Learning- Study the Joint/Posterior Probability Distribution of Dark Matter Halo simulations with Normalizing flows.- https://arxiv.org/pdf/2103.13932.pdf