Hi, I'm Nikhil
Senior Research Fellow in Computer Science, TCG CREST, India
About
I'm a final-year PhD student in Computer Science in the AI-ML group at the Institute for Advancing Intelligence (IAI), TCG CREST, and an applied mathematician by training. I work on speaker diarization, a challenging research problem in which I apply deep learning and signal processing techniques to multi-speaker conversational speech recorded in noisy, real-life conditions. My broader research interests include unsupervised machine learning and explainability through causal inference.
I am jointly supervised by Swami Punyeshwarananda (RKMVERI, Belur) and Dr. Md Sahidullah (TCG CREST).
Before starting my PhD, I completed a B.Sc. (Hons.) in Mathematics at the University of Delhi and an M.Sc. in Applied Mathematics at the Indian Institute of Technology (IIT) Mandi. During my master's, I spent a semester as a DAAD scholar at the Communications Theory Lab, TU Dresden, Germany, where I worked on an ill-posed inverse problem in image processing.
My hobbies include teaching applied mathematics, playing soccer, trekking, and practising yoga and meditation.
Currently, I am actively looking for industrial research opportunities across the broad applications of speech and audio AI.
Research interests
- Speaker diarization
- Deep learning for speech and audio
- Deep clustering
- Clustering
- Causal inference
Education
- Ph.D. in Computer Science, TCG CREST, Kolkata (2022 – 2027, expected)
- M.Sc. in Applied Mathematics, Indian Institute of Technology (IIT) Mandi (2020)
- B.Sc. (Hons.) in Mathematics, Dyal Singh College, University of Delhi (2016)
Experience
- Project Intern, Indian Institute of Technology Mandi (2020)
- Exchange Student, DAAD Scholar, Technische Universität Dresden, Germany (2018 – 2019)
Publications
Conference papers
-
Self-Tuning Spectral Clustering for Speaker Diarization
N. Raghav, A. Gupta, M. Sahidullah, S. Das
IEEE ICASSP 2025, Hyderabad, India PDF DOI Code
@inproceedings{raghav2025self, title = {Self-Tuning Spectral Clustering for Speaker Diarization}, author = {Raghav, Nikhil and Gupta, Avisek and Sahidullah, Md and Das, Swagatam}, year = {2025}, booktitle = {Proc. ICASSP}, } -
Exact Rotation Invariant Robust PCA
T. Jana, N. Raghav, A. Majumdar, M. Sahidullah
IEEE ICASSP 2025, Hyderabad, India DOI
@inproceedings{tanmoy2025exact, title = {Exact Rotation Invariant Robust PCA}, author = {Jana, Tanmoy and Raghav, Nikhil and Majumdar, Angshul and Sahidullah, Md}, year = {2025}, booktitle = {Proc. ICASSP}, } -
Assessing the Robustness of Spectral Clustering for Deep Speaker Diarization
N. Raghav, M. Sahidullah
IEEE INDICON 2024, Kharagpur, India PDF DOI
@inproceedings{raghav2024assessing, title = {Assessing the Robustness of Spectral Clustering for Deep Speaker Diarization}, author = {Raghav, Nikhil and Sahidullah, Md}, year = {2024}, booktitle = {Proc. INDICON}, }
Preprints and system descriptions
- DiariZen Explained: A Tutorial for the Open Source State-of-the-Art Speaker Diarization Pipeline N. Raghav 2026 arXiv:2604.21507 PDF Code
- TCG CREST System Description for the DISPLACE-M Challenge N. Raghav, M. Sahidullah 2026 arXiv:2603.02030 PDF
- MK-SGC-SC: Multiple Kernel Guided Sparse Graph Construction in Spectral Clustering for Unsupervised Speaker Diarization N. Raghav, A. Gupta, S. Das, M. Sahidullah 2026 arXiv:2601.19946 PDF Code
- The TCG CREST–RKMVERI Submission for the NCIIPC Startup India AI Grand Challenge N. Raghav, A. Banerjee, J. Chakraborty, A. Gupta, S. Punyeshwarananda, M. Sahidullah 2025 arXiv:2512.11009 PDF
- TCG CREST System Description for the Second DISPLACE Challenge N. Raghav, S. Saha, M. Sahidullah, S. Das 2024 arXiv:2409.15356 PDF
Challenges
- DISPLACE-M Challenge, organised by IISc Bangalore; ranked 5th on the Phase I leaderboard; system description (2026)
- NCIIPC Startup India AI Grand Challenge, team lead of a six-member team on multilingual speaker identification, diarization, transcription and translation; system description (2025)
- Second DISPLACE Challenge, TCG CREST team submission; system description (2024)
Awards
- TCG CREST PhD Fellowship (2022 – 2027)
- LDC Data Scholarship, Linguistic Data Consortium, University of Pennsylvania; data worth US$600 (Spring 2024)
- DAAD Scholarship for a semester exchange at TU Dresden, Germany (2018 – 2019)
Talks
- Causal Insights into Speaker Diarization. Invited talk, Department of Computer Science, RKMVERI, Belur (September 2026)
Teaching and service
- Teaching assistant for the Machine Learning (2025) and Deep Learning (2026) courses at TCG CREST
- Mentor to student interns at TCG CREST
- Reviewer for IEEE Transactions on Multimedia and for Computer Speech & Language
- Session chair at ICASSP 2025, in the Cybersecurity and Information Forensics track
Outreach
- AI for Everyone workshop. Resource person at a two-day workshop organised by the Department of Mathematics, Jawaharlal Nehru College, Pasighat, Arunachal Pradesh, under the DBT-STAR College Scheme, in March 2025. I delivered a lecture on AI to undergraduate mathematics and physics students and faculty. Event page
- Speech Pitch podcast. Student volunteer with the Speech Pitch team at ISCA-SAC, as a host, script writer and in post-production (sound editing and mixing)
- Vigyan Bharat. A YouTube channel on mathematics that I started during the pandemic. Making it has shaped how I think about explaining ideas clearly. Watch on YouTube
Contact
Institute for Advancing Intelligence, TCG CREST, India
nikhil.raghav.92@tcgcrest.org
raghav.nikhil29@gmail.com