Indranil Halder is a Postdoctoral Fellow at the Harvard John A. Paulson School of Engineering and Applied Sciences in the Machine Learning Foundations Group with Cengiz Pehlevan. Previously, Indranil was the Harvard Quantum Initiative Fellow at the Center for the Fundamental Laws of Nature with Daniel Jafferis and briefly served as a Research Associate at Harvard University. During his graduate studies, he was advised by Shiraz Minwalla at Tata Institute of Fundamental Research.

Research

Theoretical machine learning

Indranil is a theoretical machine learning researcher specializing in generative AI, with research interests in the interpretability of knowledge representations, the safety and robustness of deep learning systems, and inference-time algorithms.

He has defined and studied an analytically tractable one-step diffusion model, leading to a publication in the International Conference on Machine Learning (ICML). In the context of higher-dimensional statistics, using the method of deterministic equivalence, he has proved a theorem that presents an explicit formula for the Kullback-Leibler divergence between the generated and sampling distribution, taken to be isotropic Gaussian, showing the effect of finite diffusion time and noise scale. It shows that the monotonic fall phase of Kullback-Leibler divergence begins when the training dataset size reaches the dimension of the data points. Finally, for large-scale practical diffusion models, it explains why a higher number of diffusion steps enhances production quality.

In parallel, he is developing a safety-aware pre-training algorithm for large language models designed to enhance robustness while preserving performance. He also has an upcoming paper that establishes a formal inference-time scaling law and identifies the optimal number of inference-time samples for highly accurate and comparatively noisy reward processes, respectively.

Theoretical physics

Before transitioning to machine learning, Indranil made notable contributions to theoretical physics, with several publications in the Journal of High Energy Physics (JHEP).

During his early days at Harvard, Indranil Halder established a new strong-weak duality closely related to ER=EPR - the fascinating connection between entanglement and geometry. Using the duality he has made distinct progress on the long-standing issue of thermal microstate counting of blackholes. He has also proposed a framework to evaluate the supersymmetric index in disordered systems in terms of bi-local fields closely related to wormhole-like physics.

During his graduate studies, Indranil worked extensively on theoretical developments of topological quantum computation, more precisely on Chern-Simons gauge theory coupled to matter. He discovered a condensed phase and pointed out an exact non-commutative structure in the presence of a background magnetic field.