CV

Education

• Ph.D. in Computer Science, University of California, Riverside, Sep 2024 – Aug 2028 (expected)
  • Advisor: Yue Dong
  • Research focus: multimodal interpretability and safety, long-context safety, AI for science
• B.Sc. in Computer Science and Engineering, Bangladesh University of Engineering and Technology, Apr 2019 – May 2024
  • CGPA: 3.98/4.00

Research & Industry Experience

• Jun 2026 – Sep 2026: Applied Scientist Intern, Amazon (Account Integrity – Buyer Risk Prevention), San Diego, CA
  • Studied the perceptual cost of diffusion-robust adversarial perturbations; work resulted in an internal safety research publication.
• Jun 2025 – Sep 2025: Applied Scientist Intern, Amazon (Account Integrity – Buyer Risk Prevention), San Diego, CA
  • Developed CAPTCHA recommendation systems with test-time adaptation for at-gate bot mitigation in production.
• Jun 2023 – Aug 2024: Research Assistant, Samsung R&D Bangladesh
  • Developed LELANTE, an LLM-based agent for executing natural-language instructions across Android applications; the work appeared at EASE 2025.

Selected Publications

• VLMs Need Words: Vision-Language Models Ignore Visual Detail in Favor of Semantic Anchors, COLM 2026.
• AbFlowNet: Optimizing Antibody-Antigen Binding Energy via Diffusion-GFlowNet Fusion, NeurIPS 2026. Equal contribution.
• Do Reasoning LLMs Refuse What They Infer in Long Contexts?, NeurIPS 2026.
• D-Judge: Disrupting Multi-Turn Jailbreaks using Semantics-Preserving Output Rewriting, ICML 2026.
• Harnessing the Unseen: The Hidden Influence of Intrinsic Knowledge in Long-Context Language Models, AAAI 2026.
• ExpertGenQA: Open-ended QA Generation in Specialized Domains, Findings of EMNLP 2025.
• Leveraging Complementary Attention Maps in Vision Transformers for OCT Image Analysis, ICIP 2025. Equal contribution.
• IllusionVQA: A Challenging Optical Illusion Dataset for Vision-Language Models, COLM 2024. Equal contribution.
• Asymmetric Bias in Text-to-Image Generation with Adversarial Attacks, Findings of ACL 2024.
• BiRNA-BERT Allows Efficient RNA Language Modeling with Adaptive Tokenization, Nature Communications Biology. Equal contribution.

Awards

• Eastern European Machine Learning Summer School Best Poster, 2023
• IEEE Video and Image Processing Cup Champion, 2023

Talks

• Co-presenter, ACL 2024 Tutorial: Vulnerabilities of Large Language Models to Adversarial Attacks

Technical Skills

PyTorch, CUDA, Triton, JAX, vLLM, SGLang, DSPy, AWS AI Services

Full CV (PDF)