🙋♂️ Hello there, I’m Vaibhav!
Senior AI Data Scientist at MOFSL | Ph.D. Scholar at CSRE, IIT Bombay | Ex-C-MInDS M.S.
I am a Senior AI Data Scientist at Motilal Oswal Financial Services Ltd. (MOFSL) and a Ph.D. Scholar at the Centre of Studies in Resources Engineering (CSRE), IIT Bombay. I completed my M.S. by Research in AI & Data Science at C-MInDS, IIT Bombay in 2025.
My work spans machine learning, computer vision, natural language processing, and generalizable recognition systems. In industry, I build production AI systems for lip synchronization, Indic-language tokenization, and financial time-series modeling. My research focuses on models that adapt across visual domains, discover unseen categories, and remain robust in open-world settings.
🔬 Research Overview
My research asks:
How can we build intelligent systems that generalize to new domains, discover new categories, and preserve meaningful structure in data?
My core interests include:
- Domain Generalization (DG) and Generalized Category Discovery (GCD)
- Fine-grained and open-world recognition
- Hyperbolic and geometry-aware representation learning
- Vision Transformers, generative models, and Gaussian Splatting
- Indic NLP and morphology-aware tokenization
My work has appeared at CVPR 2025, NeurIPS 2025, Interspeech 2026, and ACM Transactions on Computing for Healthcare.
📝 Selected Publications
SuTRA: Structurally-Unified Tokenization with Root Awareness
Interspeech 2026 — Accepted
Vaibhav Rathore, Siddhant Gole, Dadhichi Telwadkar, Rooshil Bhatia, Maulik Ruparel, Siddharth Surekha, Neha Bhargava
A morphology-aware tokenizer that preserves Indic akshara structure and reduces morphological shattering across Hindi, Marathi, and Gujarati.
M³ QuestionIng: Multi-modal Multi-span Medical Question Answering
ACM Transactions on Computing for Healthcare — Accepted
Anisha Saha, Vaibhav Rathore, Abhisek Tiwari, Akash Ghosh, Sai Ruthvik Edara, Sriparna Saha
Introduces M³QAFrame and a medical QA dataset in which answers draw jointly from multiple textual and visual spans.
HiDISC: A Hyperbolic Framework for Domain Generalization with Generalized Category Discovery
NeurIPS 2025
Vaibhav Rathore, Divyam Gupta, Biplab Banerjee
A hyperbolic DG-GCD framework that improves cross-domain alignment and novel-category discovery without expensive episodic simulation.
Paper · Code · Project · Slides · Poster
When Domain Generalization Meets Generalized Category Discovery
CVPR 2025
Vaibhav Rathore, Shubhranil B, Saikat Dutta, Sarthak Mehrotra, Zsolt Kira, Biplab Banerjee
Introduces DG²CD-Net and the DG-GCD setting, using episodic adaptation and adaptive task arithmetic to generalize to unseen domains and categories.
Paper · Code · Project · Presentation · Poster
Under Review
FOCUS: Bridging Fine-Grained Recognition and Open-World Discovery across Domains
Under review at WACV 2027
Vaibhav Rathore, Divyam Gupta, Moloud Abdar, Subhasis Chaudhuri, Biplab Banerjee
Preprint
💼 Experience
Senior AI Data Scientist — Motilal Oswal Financial Services Ltd. (MOFSL)
Nov 2025–Present
- Member of the AI Research Team, working on applied and foundational AI research for the Indian context.
- Researching time-series and tabular foundation models for financial data.
- Developing Indic TTS/ASR tokenization and production-grade lip-sync systems.
- Actively developing research for submission to top-tier AI and machine-learning venues.
Public descriptions of this work are intentionally high-level because the underlying MOFSL projects and results are proprietary.
Research Intern — Motilal Oswal Financial Services Ltd. (MOFSL)
Sep 2025–Oct 2025
- Evaluated LatentSync, OmniSync, KeySync, and Gaussian Splatting for high-fidelity lip synchronization and 3D avatar generation.
Research Intern — Sony Research India
May 2025–Jul 2025
- Developed a 3D lip-sync pipeline using Gaussian Splatting and controllable 3D avatars.
Research Intern — Vehant Technologies
Sep 2025–Nov 2025 · Vehant Fellow Track
- Researched Pedestrian Attribute Recognition using hyperbolic embeddings, attribute graphs, correlation-aware models, and vision-language models.
- Studied robustness to occlusion and domain shift, along with fairness across PAR benchmarks.
Research & Development Intern — Clinical AI Assistance
May 2024–Jul 2024
- Fine-tuned LLMs with LoRA and developed a multimodal medical QA system combining textual and imaging information.
Graduate Engineer Trainee — Reliance Industries Ltd.
Aug 2022–Jul 2023
- Developed predictive-maintenance and operational-optimization models.
🎓 Education
Ph.D. Scholar, CSRE, IIT Bombay
2026–Present · CPI: 9.00
M.S. by Research, AI & Data Science, C-MInDS, IIT Bombay
2023–2025 · Completed · CPI: 9.31
B.Tech., Mechanical Engineering, MNNIT Allahabad
2018–2022 · CPI: 9.10
🏆 Awards & Honors
- Award for Excellence in Research, M.S. by Research, C-MInDS, IIT Bombay — August 22, 2026
- Top 1 percentile, JEE Main 2018 (among 1.2M candidates)
- GATE 2022: 97.84 percentile in Engineering Sciences (XE) and 96 percentile in Mechanical Engineering (ME)
🧑🏫 Teaching Experience
Teaching Assistant, IIT Bombay
- DS 303: Introduction to Machine Learning
- e-PGD: Python Programming
🛠️ Technical Skills
Programming: Python, C++, C, Java, Bash
ML/DL: PyTorch, TensorFlow, scikit-learn, OpenCV, YOLO
Generative AI: VAE, GANs, Gaussian Splatting, lip synchronization
NLP & Data: Hugging Face, Indic tokenization, Pandas, NumPy, financial time series
Applications: Streamlit, production ML pipelines, data visualization
📫 Contact
I am open to collaborating with independent researchers on exciting research projects. If our interests overlap, feel free to email me with a brief introduction and project idea; we can explore working together at mutually suitable times.
📍 Mumbai, India
📧 vaibhav.rathor.in@gmail.com
🔗 Google Scholar · GitHub · LinkedIn · ORCID · OpenReview