Ambarish Moharil

Ambarish Moharil MSc

PhD Candidate at SAFEGUARD lab

• Research profile: Geometric Deep Learning and XAI
• Research expertise: Automated Machine Learning/Deep Learning, Machine/Deep Learning for Spatio-Temporal Multimodal Data, Model-Based Bayesian Optimization, Advanced Statistical Analysis.

Biography

Ambarish Moharil is a PhD Candidate at Safeguard Lab, working with the Data Governance unit at JADS in collaboration with Deloitte. His research focuses on the explainability and transparency of deep learning models, with a specific interest in exploring geometric symmetries in neural parameter spaces, such as scale-invariance. Ambarish is developing methods to analyze dynamic representations of neural networks in high-dimensional spaces while retaining important symmetries like permutation invariance. He has published extensively in top-tier venues such as ICSE-MSR, NLBSE, ECML PKDD, and KDD. Recently, he designed a hybrid AutoDL tool for synthesizing multimodal pipelines using pre-trained transformers and created the TABASCO toolkit to assist in disambiguating amorphous nouns in software requirements documents. His research aims to enhance human-AI collaboration by examining the impact and behavior of AI systems.

Relevant academic links

LinkedIN Google Scholar