Ravi Ranjan | Trustworthy AI Researcher, PhD Candidate at FIU

Hi, I’m Ravi. I am a Computer Science Ph.D. candidate at Florida International University (FIU), advised by Prof. Agoritsa Polyzou. I’m interested in what happens after an AI model is trained: what it remembers, how its behavior changes, and how we can correct it without losing what still works.

My research sits at the intersection of trustworthy foundation models, machine unlearning, privacy auditing, fairness, explainability, and post-training adaptation. My first-author work has appeared at CVPR, ICPR, IJCNN, ACL, AACL-IJCNLP, and INTERSPEECH, spanning language, vision, diffusion, audio, and embodied AI. I have also been fortunate to receive FIU’s Doctoral Evidence Acquisition Fellowship and an earlier research fellowship for applying machine learning to cryptocurrency forecasting.

My path to research has not been entirely academic. Before and alongside my Ph.D., I have worked across AI, data systems, cloud computing, and software engineering, and taught and mentored AI/ML learners through Cornell University (Break Through Tech). That mix of research and engineering shapes how I think about AI: I like ideas that do more than work on paper—they should be testable, reproducible, and useful in the real world.

I’m always happy to exchange ideas about trustworthy AI, research collaborations, or interesting problems worth solving. Feel free to reach out through LinkedIn or email.

Events & News

[2026/09] Our paper Forgetting Only What Matters: Layer-Selective Unlearning toward Robust LLMs was accepted to AACL-IJCNLP 2026 for an oral presentation.
[2026/08] Concluded my role as an AI & ML Instructor with Cornell University’s Break Through Tech (BTT) , where I supported learners in machine learning, Python, applied AI, and hands-on project development.
[2026/08] New preprint online: SafeCommit: Certifying When Memory-Grounded Agents May Safely Act .
[2026/07] Awarded the Doctoral Evidence Acquisition (DEA) Fellowship from the FIU University Graduate School , supporting my doctoral research and dissertation progress.
[2026/07] Served as a Session Chair at ACL 2026, chairing a session on safety and alignment in large language models.
[2026/07] Virtually attended ACL 2026 to present two research papers in poster sessions.
[2026/06] Served as a Session Chair at IJCNN 2026, chairing sessions on Engineering Trust: Ethical, Legal, and Societal Impacts of Computational Intelligence on Human Agency and Explainability and Security in Trustworthy Artificial Intelligence Systems.
[2026/06] Virtually attended IJCNN 2026 for the oral presentation of our paper.
[2026/06] Attended CVPR 2026 in Denver and presented our work in the poster session.
[2026/06] Received an ACL 2026 Registration Grant .
[2026/06] Our paper Listening with Attention: Entropy-Guided Explainability for Transformer-Based Audio Models was accepted to INTERSPEECH 2026 for an oral presentation.
[2026/05] Successfully defended my Ph.D. dissertation proposal at FIU.
[2026/04] Our paper PERSA: Reinforcement Learning for Professor-Style Personalized Feedback with LLMs was accepted to the 21st Workshop on Innovative Use of NLP for Building Educational Applications (BEA) at ACL 2026.
[2026/04] Our paper VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Models was accepted to the 4th Workshop on Towards Knowledgeable Foundation Models (KnowFM 2026) at ACL 2026.
[2026/04] Received a CVPR 2026 Travel Grant to support conference participation.
[2026/04] Our paper G-Drift MIA: Membership Inference via Gradient-Induced Feature Drift in LLMs was accepted to ICPR 2026.
[2026/03] Our paper CatRAG: Functor-Guided Structural Debiasing with Retrieval Augmentation for Fair LLMs was accepted to IJCNN 2026 and selected as a Best Paper Nominee.
[2026/03] New preprint online: Position: LLMs Must Use Functor-Based and RAG-Driven Bias Mitigation for Fairness .
[2026/03] Our paper RAZOR: Ratio-Aware Layer Editing for Targeted Unlearning in Vision Transformers and Diffusion Models was accepted to the CVPR 2026 Findings track.
[2026/03] New preprint online: Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies .
[2025/11] Passed the Ph.D. qualifying examination at FIU and officially advanced to Ph.D. candidacy.
[2024/12] New preprint online: Trustworthiness of LLMs in Medical Domain .
[2024/08] Started my Ph.D. in Computer Science at Florida International University (FIU) .
[2023/05] Graduated from Rowan University with an M.S. in Data Science.
[2023/03] My student research paper Effectiveness of Machine Learning Algorithms in Cryptocurrency Price Forecasting was published through the Phemex Student Foundation.
[2023/02] Proud to be part of Rowan University's SEM team, which received a NASPA Award for improving and simplifying the admissions process.
[2022/08] Published a Phemex article, Retailers Planning to Accept Crypto Payments .
[2022/03] Selected for the Phemex Research Fellowship in Machine Learning and Cryptocurrency Forecasting , as the sole recipient from Rowan University.

RAZOR: targeted unlearning in vision transformers and diffusion models

RAZOR: Ratio-Aware Layer Editing for Targeted Unlearning in Vision Transformers and Diffusion Models [PDF] [Code]

Ravi Ranjan¹, Utkarsh Grover, Xiaomin Lin, Agoritsa Polyzou

CVPR 2026 Findings

Machine Unlearning Vision Transformers Diffusion Models Selective Editing Trustworthy AI

G-Drift MIA: membership inference in large language models

G-Drift MIA: Membership Inference via Gradient-Induced Feature Drift in LLMs [PDF]

Ravi Ranjan¹, Utkarsh Grover, Xiaomin Lin, Agoritsa Polyzou

ICPR 2026

Membership Inference LLM Privacy Gradient Drift Privacy Auditing Large Language Models

CatRAG: fairness and structural debiasing for large language models

CatRAG: Functor-Guided Structural Debiasing with Retrieval Augmentation for Fair LLMs [PDF]

Ravi Ranjan¹, Utkarsh Grover, Mayur Akewar, Xiaomin Lin, Agoritsa Polyzou

IJCNN 2026 · Best Paper Nominee

Fairness RAG LLM Debiasing Category Theory Responsible AI

VLA-Forget: unlearning in vision-language-action models

VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Models [PDF] [Code]

Ravi Ranjan¹, Agoritsa Polyzou

ACL 2026 · KnowFM

VLA Unlearning Embodied AI Robot Learning Safety Selective Unlearning

PERSA: personalized feedback with reinforcement learning and large language models

PERSA: Reinforcement Learning for Professor-Style Personalized Feedback with LLMs [PDF]

Ravi Ranjan¹, Utkarsh Grover, Xiaomin Lin, Agoritsa Polyzou

ACL 2026 · BEA

LLM Personalization Reinforcement Learning AI in Education Personalized Feedback Large Language Models

Listening with Attention: explainability for transformer-based audio models

Listening with Attention: Entropy-Guided Explainability for Transformer-Based Audio Models [PDF] [Code]

Ravi Ranjan¹, Utkarsh Grover, Xiaomin Lin, Agoritsa Polyzou

INTERSPEECH 2026 · Oral Presentation

Explainable AI Audio Transformers Attention Entropy Speech AI

Forgetting Only What Matters: layer-selective unlearning for large language models

Forgetting Only What Matters: Layer-Selective Unlearning toward Robust LLMs [PDF]

Ravi Ranjan¹, Olivera Kotevska, Agoritsa Polyzou

AACL-IJCNLP 2026 · Oral Presentation

LLM Unlearning Layer-Selective Editing Quantization Robustness Privacy Trustworthy LLMs

All research papers and publications.

SafeCommit: Certifying When Memory-Grounded Agents May Safely Act [PDF]

Mayur Akewar¹, Ravi Ranjan¹

Preprint · 2026

AI Agents Agent Safety Memory-Grounded Agents Trustworthy AI

Survey of embodied foundation models at the edge

Embodied Foundation Models at the Edge: A Survey of Deployment Constraints and Mitigation Strategies [PDF]

Utkarsh Grover¹, Ravi Ranjan², Mingyang Mao, Trung Tien Dong, Satvik Praveen, Zhenqi Wu, J Morris Chang, Tinoosh Mohsenin, Yi Sheng, Agoritsa Polyzou, Eiman Kanjo, Xiaomin Lin

Preprint · 2026

Embodied AI Foundation Models Edge AI Deployment AI Safety

Position paper on functor-based and RAG-driven bias mitigation

Position: LLMs Must Use Functor-Based and RAG-Driven Bias Mitigation for Fairness [PDF]

Ravi Ranjan¹, Utkarsh Grover, Agoritsa Polyzou

Preprint · 2026

LLM Fairness RAG Bias Mitigation Category Theory Responsible AI

Trustworthiness of large language models in the medical domain

Trustworthiness of LLMs in Medical Domain [PDF]

Ravi Ranjan¹, Vishal Pramanik, Utkarsh Grover, Venkata Ramesh Ganapam

Research Preprint · 2024

Medical AI LLM Trustworthiness Explainable AI Responsible AI

Awards

[2026/08] Doctoral Evidence Acquisition (DEA) Fellowship — University Graduate School, Florida International University ($13,136.54 + tuition waiver).
[2026/07] ACL 2026 Conference Registration Grant ($600).
[2026/06] CVPR 2026 Conference Travel Grant ($1,000 + registration waiver).
[2026/03] ★ Outstanding Reviewer Award — ICML Mechanistic Interpretability Workshop.
[2023/02] Part of Rowan University’s SEM team recognized with a NASPA Award for improving and simplifying the university admissions process.
[2022] Academic Excellence Recognition — Rowan University, for achieving a Top 5% GPA.
[2022] Phemex Research Fellowship Award — selected as the sole Rowan University student recipient ($3,000).

Academic Service

Reviewer

[2026] NeurIPS — Ethics Review, Conference & Datasets/Benchmarks Track
[2025, 2026] AAAI Conference on Artificial Intelligence
[2026] Association for Computational Linguistics (ACL)
[2026] Educational Data Mining (EDM)
[2026] IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
[2026] International Joint Conference on Neural Networks (IJCNN)
[2026] ICML, Mechanistic Interpretability Workshop ★ Outstanding Reviewer Award

Session Chair

[2026] ACL 2026 — Session on Safety and Alignment in Large Language Models
[2026] IJCNN 2026 — Chaired sessions on Engineering Trust: Ethical, Legal, and Societal Impacts of Computational Intelligence on Human Agency and Explainability and Security in Trustworthy Artificial Intelligence Systems.