Speech

Most of my work has been conducted at Sony Research India.

• Automatic Speech Recognition (ASR), Speaker Diarization (SD) and Voice Activity Detection (VAD) in Indian languages
• Disentangling speaker and language information from speech
• Efficient replacement for self-attention in Transformer models for speech

Language

Personal interest and ongoing research.

• LLMs for low-resource languages using Parameter-efficient Transfer Learning
• Interpretability of LLMs
• Cross-lingual Information Retrieval
• Evaluation on multilingual datasets

Tabular Data

Extremely fun and easy to train, can't complain.

• Interpretable models for tabular data
• Self-supervised learning for Anomaly Detection
• Can deep learning even beat XGBoost? In this lifetime?