Shubhashis Roy Dipta

PhD ResearcherReasoningMultimodal, NLP

prof_pic.jpg

Shubhashis Roy Dipta

sroydip1@umbc.edu


Affiliations

Amazon Science
Applied Research Scientist Intern
Summer 2025
Manager: Dr. Lichao Wang
Mentor: Dr. Daniel Bis
Scale.AI
Machine Learning Research Intern
Summer 2024
Manager: Dr. Adrian Lam
Mentor: Vijay Kalmath
University of Maryland, Baltimore County
Ph.D. in Computer Science
Spring 2021 - Present
Advisor: Dr. Frank Ferraro
Grade: 4.00/4.00
Publications: See Here (From 2023)
University of Maryland, Baltimore County
M.Sc. in Computer Science
Spring 2021 - Spring 2023
Awards: Phi Kappa Phi
Grade: 4.00/4.00
Morgan State University
Research Assistant
2017 - 2019
Advisor: Dr. Iman Dehzangi
Publications: 4 Journals (Genes, Elsevier, IEEE Access, Springer)
UniShopr.com
Founder
2017 - 2021
Military Institute of Science & Technology
B.Sc. in Computer Science
Spring 2013 - Fall 2017
Advisor: Dr. Wali Mohammad Abdullah
Grade: 3.51/4.00

Professional Services

Reviewer of Position Paper at NeurIPS 2025
Reviewer of ELVM workshop at CVPR 2025
Reviewer of ACL 2025 (6 papers)
Reviewer of W-NUT workshop at NAACL 2025
Reviewer of TrustNLP Workshop at NAACL 2025 (2 Papers)
Reviewer of NAACL Industry Track 2025 (1 Paper)
Reviewer of COLING 2025 (2 Papers)
Reviewer of BMC Bioinformatics (July 2024)
Reviewer of Scientific Reports (July 2024)
Reviewer of SRW at NAACL 2024 (2 Papers)
Reviewer of SemEval-2024 (4 papers)
Reviewer of Scientific Reports, Nature (Jan 2024)
Reviewer of Plant Methods (Jan 2024)
Reviewer of W-NUT workshop at EACL 2024
Reviewer of Plant Methods (Dec 2023)
Reviewer of Computational and Structural Biotechnology Journal (Mar 2023)
Secondary Reviewer of *SEM 2023

Shubhashis is a Computer Science PhD Researcher under Dr. Frank Ferarro at the University of Maryland, Baltimore County (UMBC). His research combines Natural Language Processing (NLP) and Computer Vision (CV).

His broad research focuses on Decomposition-based Reasoning using text or vision data. Currently, he is focused on question based decomposition to understand the scientific feasibility of a given task.

Over the years, he has worked on Outcome and Intention based Decomposition and how it can be used for video-text retrieval. This research has applications in video/image retrieval system, especially where there is no text metadata available for the video (e.g., most of the videos on the internet, social media, and surveillance videos). Also, his previous work, a hierarchical variational autoencoder for Event Representation Learning, has applications in text summarization, question answering, and counterfactual reasoning (Published in *SEM 2023, ACL).

In Summer 2025, he interned at Amazon Science as a Applied Scientist. He has shown how Deliberative Alignment can be used to optimize cost by ruducing token of the tool-calling LLMs (paper coming soon). He was mentored by Dr. Daniel Bis and managed by Dr. Lichao Wang.

In Summer 2024, he interned at Scale.AI as a Machine Learning Researcher. He explored how RLHF can improve the text2SQL generation (currently under ARR review). He also worked on Many-Shot text2SQL and text2SQL AutoEval using SLM. He was mentored by Vijay Kalmath and managed by Dr. Adrian Lam.

If you are interested in collaborating, please email me with a short description of your research interest.

Shubhashis has a strong background in ML programming, including PyTorch, 🤗 HuggingFace, NLTK, Spacy, Matplotlib, Seaborn, and more. He has excelled in machine learning competitions (Kaggle top-70 🥉) and coding competitions (ACM ICPC 8th out of 300+ teams) and more. He was the founder of UniShopr (2018-2022), a cross-border e-commerce for his home country (Bangladesh).

Research Interest

        ✓ Math Reasoning
        ✓ Decomposition-based Reasoning
        ✓ Natural Language Understanding

Recent News

Sep 23, 2025 🥳 New Pre-print Alert: Advancing Reference-free Evaluation of Video Captions with Factual Analysis
Sep 5, 2025 🥳 Paper accepted at *SEM 2025: If We May De-Presuppose: Robustly Verifying Claims through Presupposition-Free Question Decomposition
Sep 4, 2025 🥳 New Pre-print Alert: FedMentor: Domain-Aware Differential Privacy for Heterogeneous Federated LLMs in Mental Health
Jun 14, 2025 🥳 New Pre-print Alert: Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
May 19, 2025 Started my summer internship at Amazon in Alexa Conversational AI team. I am working on optimizing the Alexa conversation latency, cost and performance.

WRITE ✍️  on Machine Learning, NLP, Vision, Multimodal AI

Featured Publications

Check out Google Scholar for a full list of my publications.

  1. EMNLP
    If We May De-Presuppose: Robustly Verifying Claims through Presupposition-Free Question Decomposition
    Shubhashis Roy Dipta, and Francis Ferraro
    In Proceedings of the 14th Joint Conference on Lexical and Computational Semantics (*SEM 2025), Jul 2025
  2. Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
    Shubhashis Roy Dipta, and Francis Ferraro
    Jul 2025
  3. Semantically-informed Hierarchical Event Modeling
    Shubhashis Roy DiptaMehdi Rezaee, and Francis Ferraro
    In Proceedings of the 12th Joint Conference on Lexical and Computational Semantics (*SEM 2023), Jul 2023