Sidharth Nair

Hello there! I am a machine learning researcher working across generative models, embodied AI, and visual computing. My Master’s thesis work at the Max Planck Institute for Informatics in Saarbruecken focuses on diffusion-driven physics-based humanoid motion and stability.

Previously, I completed my Bachelor’s in Electrical and Electronics Engineering at BITS Pilani and my Bachelor’s thesis at the Indian Institute of Science under Prof. Sundeep Chepuri, working on unsupervised neural precoding for integrated sensing and communication systems.

CV Email GitHub Scholar

research interests

Generative models for embodied AI and motion synthesis
Visual computing, 3D scene understanding, and neural rendering
LLM-driven systems for structured reasoning and human-in-the-loop workflows

news

Dec 1, 2025	Our work on AuditCopilot appeared at the NeurIPS 2025 Workshop on Generative AI in Finance.
Oct 1, 2025	Started my Master’s thesis at the Max Planck Institute for Informatics.
Jul 15, 2024	Started as a research assistant at the Interactive Machine Learning Lab, DFKI.
Oct 22, 2022	Joined the Visual Computing Master’s program at Saarland University.
Jun 12, 2022	Accepted into the LOGML Summer School, 2022!

selected publications

NeurIPS Workshop

AuditCopilot: Leveraging LLMs for Fraud Detection in Double-Entry Bookkeeping

Md Abdul Kadir, Sidharth S. Nair, S. S. M. Vasu, and 1 more author

2025

Abs arXiv PDF

AuditCopilot is a hybrid LLM-driven anomaly detection system for accounting data, combining learned anomaly scoring with rule-based audit logic and natural language explanations for model decisions.
EUSIPCO

Learning to Precode for Integrated Sensing and Communication Systems

RS Prasobh Sankar, Sidharth Nair, Siddhant Doshi, and 1 more author

2023

Abs arXiv Code

In this paper, we present an unsupervised learning neural model to design transmit precoders for integrated sensing and communication (ISAC) systems to maximize the worst-case target illumination power while ensuring a minimum signal-to-interference-plus-noise ratio (SINR) for all the users. The problem of learning transmit precoders from uplink pilots and echoes can be viewed as a parameterized function estimation problem and we propose to learn this function using a neural network model. To learn the neural network parameters, we develop a novel loss function based on the first-order optimality conditions to incorporate the SINR and power constraints. Through numerical simulations, we demonstrate that the proposed method outperforms traditional optimization-based methods in presence of channel estimation errors while incurring lesser computational complexity and generalizing well across different channel conditions that were not shown during training."