Debdeep Sanyal
Research Engineer — Founding Team
Birla AI Labs
Hello. I like computers (adult fascination, really) and I also happen to have studied computer science for my bachelors, which is great. I'm still figuring out my interest niche, but I have noticed that I'm always a bit more inclined towards the fundamentals of things, be it computer systems or neural networks.
My research spectrum has been broad and not in a bragging way. When I had started out, I had no reading taste and just read about anything which came up. If something seemed pretty off about some work, I tried fixing it. That's really how I got my initial papers. I like reading, and lately working, on works which questions core concepts, breaks assumption walls, and tries to approach things from first principles.
I enjoy building foundation models at my work, working with large systems enthuse me. I've also learnt that training models is more of an intuition and observation art which I thoroughly enjoy. I am looking for a PhD, among all the recent speculations of the worth of doing a PhD. I'll do it for the fun of doing hard things.
This entire text is composed of human generated tokens and the stochasticity is intentional, for beyond computers and maths, I am a whole human.
Education
KIIT University
B.Tech in Computer Science and Engineering
Publications
EMNLP Main 2026
Policy Optimization Prefers The Path of Least Resistance
Debdeep Sanyal, Aakash Sen Sharma, Dhruv Kumar, Saurabh Deshpande, Murari Mandal
ICML 2026
The Realignment Problem: When Right becomes Wrong in LLMs
Aakash Sen Sharma, Debdeep Sanyal, Manodeep Ray, Vivek Srivastava, Shirish Karande, Murari Mandal
ICML Workshop on Foundation Models for Structured Data (FMSD) 2026
Mix, Don't Pick: Why Synthetic Corpus Composition Matters for Time Series Foundation Model Pretraining
Aaryan Nagpal, Debdeep Sanyal, Murari Mandal, Dhruv Kumar, Saurabh Deshpande
AAAI 2026
AntiDote: Bi-level Adversarial Training for Tamper-Resistant LLMs
Debdeep Sanyal, Manodeep Ray, Murari Mandal
We make an LLM robust against malicious fine-tuning by co-evolving it in a game against an adversary model.
EMNLP Main 2026
OrgAccess: A Benchmark for Role Based Access Control in Organization Scale LLMs
Debdeep Sanyal, Umakanta Maharana, Yash Sinha, Hong Ming Tan, Mohan Kankanhalli, Murari Mandal
We introduce a novel, expert-crafted benchmark to test if LLMs can truly understand and respect complex organizational roles and permissions for enterprise use.
EMNLP Main 2025
Investigating Pedagogical Teacher and Student LLM Agents: Genetic Adaptation Meets Retrieval Augmented Generation Across Learning Style
Debdeep Sanyal, Agniva Maiti, Umakanta Maharana, Dhruv Kumar, Ankur Mali, C. Lee Giles, Murari Mandal
We created a digital sandbox for an AI to practice teaching and discover what truly works for different students.
EMNLP Findings, EMNLP NLLP Workshop Oral 2025
Nine Ways to Break Copyright Law and Why Our LLM Won't: A Fair Use Aligned Generation Framework
Aakash Sen Sharma, Debdeep Sanyal, Priyansh Srivastava, Sundar Atreya H., Shirish Karande, Mohan Kankanhalli, Murari Mandal
LAW-LM empowers LLMs to generate maximally helpful content while thoughtfully adhering to complex copyright laws.
NeurIPS BERT2S Workshop 2025
time2time: Causal Intervention in Hidden States to Simulate Rare Events in Time Series Foundation Models
Debdeep Sanyal, Aaryan Nagpal, Dhruv Kumar, Murari Mandal, Saurabh Deshpande
We show how to take the essence of a past event, like a market crash, and directly implant it into a TSFM to change how it models the future.
Experience
Research Engineer — Founding Team — Birla AI Labs
I build time-series foundation models, owning the training pipeline, experimentation, evaluation, and inference. I currently lead pretraining and architecture experimentation for Bharat-TSFM under the IndiaAI initiative.