curriculum-vitae
Basics
| Name | Mahesh Bhosale |
| Position | Computer Science PhD Candidate, University at Buffalo |
| mbhosale@buffalo.edu | |
| Url | https://bhosalems.github.io |
| Summary | PhD candidate at the University at Buffalo advised by Dr. David Doermann. Research interests: image/video synthesis via diffusion models, and VQA and fairness issues of Multi-Modal Large Language Models (MLLMs). |
Work
-
2026.06 - 2026.08 Baltimore, MD, USA
-
2022.05 - Present Buffalo, NY, USA
Research Assistant & Lab Manager
Institute of Artificial Intelligence and Data Science, University at Buffalo
Research topics: multi-modal generative AI. Advised by Dr. David Doermann (A2IL Lab).
-
2017.08 - 2021.08 Pune, MH, India
Software Engineer
Veritas Technologies LLC
Advised by Anindya Banerjee. Developed a novel scheduling algorithm on top of LSTM-based storage and compute resource prediction (IDLEBOT), reducing execution time of resource-intensive tasks by 56%. Benchmarked deep-learning-based sequence modeling solutions and built a ransomware-detection tool modeling the file change log in the Veritas Filesystem (VxFS). Bestowed Certificate of Merit, Veritas Illuminate 2020; led the effort to start an ML-for-Systems group in the org.
-
2017.01 - 2017.06 Pune, MH, India
Software Engineer Intern
Veritas Technologies LLC
Reduced execution time of variants of the 'ls' command by 20% using a directory inode read-ahead algorithm.
Education
-
2023.09 - 2027.05 Buffalo, NY, USA
-
2021.09 - 2023.05 Buffalo, NY, USA
Master of Science (by Research)
University at Buffalo, The State University of New York
Computer Science
-
2013.06 - 2017.05 Sangli, MH, India
Publications
-
2026.08.01 TRACE: Evidence Grounding-Guided Multi-Video Event Understanding and Claim Generation
ACL 2026, MAGMaR Workshop (Multimodal Augmented Generation via Multimodal Retrieval)
Pengyu Yan*, Akhil V S S Gorugantu*, Mahesh Bhosale, Abdul Wasi, Vishvesh Trivedi, David Doermann
-
2026.08.01 CRAFT: Critic-Refined Adaptive Key-Frame Targeting for Multimodal Video Question Answering
ACL 2026, MAGMaR Workshop (Multimodal Augmented Generation via Multimodal Retrieval)
Mahesh Bhosale*, Abdul Wasi*, Vishvesh Trivedi*, Pengyu Yan, Akhil V S S Gorugantu, David Doermann
-
2026.07.01 Category-wise Structured Radiology Report Generation with Contrastive Decoding
Medical Imaging with Deep Learning (MIDL) 2026 — Oral (top <= 8% of submissions)
Shantam Shrivastava, Mahesh Bhosale, David Doermann, Mingchen Gao
-
2026.06.01 FairLLaVA: Fairness-Aware Parameter-Efficient Fine-Tuning for Large Vision-Language Assistants
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026
Mahesh Bhosale, Abdul Wasi*, Shantam Shrivastava*, Shifa Latif, Tianyu Luan, Mingchen Gao, David Doermann, Xuan Gong
-
2026.01.01 RADI3N: Training-Free Geometric Image Editing via Parametric 3D Control
Under Review, 2026
Abdul Wasi, Akhil V S S Gorugantu, Mahesh Bhosale, Sauradip Nag, Anjan Dutta, Junsong Yuan, David Doermann
-
2026.01.01 Variance-Guided Score Regularization for Hallucination Mitigation in Diffusion Models
Under Review, 2026
Mahesh Bhosale*, Naresh Kumar Devulapalli*, Abdul Wasi, Chau Pham, Vishnu Lokhande, David Doermann
-
2025.12.01 AutoEdit: Automatic Hyperparameter Tuning for Image Editing
Conference on Neural Information Processing Systems (NeurIPS) 2025
Chau Pham, Quan Dao, Mahesh Bhosale, Yunjie Tian, Dimitris N. Metaxas, David Doermann
-
2025.10.01 PathDiff: Histopathology Image Synthesis with Unpaired Text and Mask Conditions
International Conference on Computer Vision (ICCV) 2025
Mahesh Bhosale, Abdul Wasi, Yuanhao Zhai, Yunjie Tian, Samuel Border, Nan Xi, Pinaki Sarder, Junsong Yuan, David Doermann, Xuan Gong
-
2024.10.01 ChartReformer: Natural Language-Driven Chart Image Editing
International Conference on Document Analysis and Recognition (ICDAR) 2024
Pengyu Yan, Mahesh Bhosale, Jay Lal, Bikhyat Adhikari, David Doermann
-
2023.08.01 LineFormer: Rethinking Line Chart Data Extraction as Instance Segmentation
International Conference on Document Analysis and Recognition (ICDAR) 2023
Jay Lal, Mahesh Bhosale*, Aditya Mitkari*, David Doermann
-
2023.01.01 Player Re-Identification Using Body Part Appearances
arXiv 2023
Mahesh Bhosale*, Abhishek Kumar*, David Doermann
Projects
- 2025.01 - Present
Id-Diff
Exploring relationships between the intrinsic dimension of data and the generative performance of latent diffusion models. In collaboration with Carez AI and the University of Montreal.
- 2025.01 - Present
AI Agents for Pathology
Exploring agentic AI for computational pathology that predicts patient survival while remaining clinically interpretable. In collaboration with NYU Langone Health.
- 2025.01 - Present
Few-Step Distillation of Flow Models
Tackling the instability of GAN training when distilling flow-based generative models into few-step samplers. In collaboration with Meta.
- 2025.01 - Present
Multi-Video Understanding
Improving the efficiency and grounding of article generation from multiple videos given a query. In collaboration with Johns Hopkins University.
Awards
- 2023.02.01
Research Mentorship Program Scholar
Google
Accepted to a three-month program that matches students with Google mentors and peers to support their pursuit of computer science research pathways.
- 2020.03.01
Certificate of Merit - Veritas Illuminate
Veritas Technologies LLC
Presented a paper titled 'IDLEBOT - Predictive execution of resource-intensive tasks' at Veritas's annual technical conference.
- 2016.04.01
Silver Medal, World Code Sprint
HackerRank
Awarded a silver medal in the World Code Sprint competitive programming contest.
- 2015.12.01
Honorable Mention, ACM-ICPC Amritapuri Regional
Association for Computing Machinery and Baylor University
Represented Walchand College of Engineering for the first time at the International Collegiate Programming Contest regionals at Amrita University, Coimbatore.
- 2015.01.01
Runner Up, Best Project Competition
Department of Information Technology, WCE, Sangli
Ranked 2nd best project in the department for computing top-10 Twitter hashtags with real-time processing and sentiment analysis using Apache Spark, Beautiful Soup, and the Twitter4j API, visualized with D3.js.
- 2014.11.01
Runner Up, MindSpark Coding Competition
College of Engineering, Pune
An algorithmic coding competition with over 500 students in teams of two from colleges across India, spanning MCQ, short-coding, and long-coding rounds.
- 2014.01.01
DST Inspire Science Camp (Top 1%)
Indian Institute of Information Technology, Pune
Selected (top 1%) for a week-long science camp hosted by the Department of Science and Technology, Government of India, and IIIT Pune.