Projects

Current projects

Predicting delay in Rail Network

Here, we study train delay prediction for express trains in Indian Rail Network. This project involves designing transformer-based prediction models in spatio-temporal time series data.

Sustainable AI

This project explores accelerating LLM inferences for energy optimization and carbon footprint reduction. Towards that, we explore various techniques like output compression, employing energy-optimal assistant models, etc.

Generative models for temporal event data

This project involves designing TPP-based diffusion models.

Explainable AI

In this project, we are working on developing domain-agnostic attribution techniques and their aplicaiton in diverse domains, namely sequencial data modeling (temporal data models and LLMs in particular), vision, and biology.

Projects in PhD tenure

Incentive design in unequal competition

Here, we address the problem of designing dynamic incentives in team-based competitive games where two teams do not have equal capabilities using a multi-agent reinforcement learning framework.

Human-assisted machine learning models

Here we address the problem of designing human-assisted machine learning models in a linear regression setup.

Demarcation of Marked Temporal Point Process

Here we explore various designing algorithms for demarcating heterogenous and homogenous messages from given streamed messages over a network using a marked temporal point process framework.