experience
Research, teaching, and competitive programming experience.
🔬 Research Experience
Research Intern — Artificial Intelligence Institute of the University of South Carolina (AIISC)
Remote · Sep 2024 – Present
Topics: Diffusion Models, Adversarial Attacks on LLMs/VLMs, Speculative Decoding
- Collaborating with Prof. Amitava Das (Professor, BITS Pilani; Ex-AIISC), Aman Chadha (GenAI Manager, AWS), and Vinija Jain (ML Lead, Meta AI) on LLMs and diffusion-model research.
- Researched hallucination detection in text-to-image (T2I) generation using multimodal models (e.g., Qwen-VL, LLaVA).
- Evaluated hallucination phenomena with MS-COCO and T2I benchmarks for semantic alignment and visual–textual consistency.
- Investigating speculative decoding–based defenses against adversarial attacks on diffusion models.
- Research Collaborator, Qatar Computing Research Institute (QCRI) …………………………………. Remote
Focus: Multimodal LLMs & Agentic AI, Spatial Reasoning, Geospatial Reasoning, and Geotemporal Reasoning
Supervised by: Md Rizwan Parvez, PhD
Jun. 2025 – Present- Develop and evaluate multimodal LLMs and VLMs for spatial reasoning and geospatial/geotemporal reasoning.
- One paper on spatial reasoning accepted at ICLR 2026.
- Lead author of one paper on geotemporal reasoning under review at ICML 2026.
- Lead author of one paper on geospatial and geotemporal reasoning under review at ACL ARR March 2026.
- Research Collaboration, Cognitive Agents and Interaction Lab, University of Dhaka …………………………………. Remote
Focus: Symbolic reasoning in LLMs and formal reasoning benchmarks
Supervised by: Dr. Md Mosaddek Khan
Dec. 2025 – Present- Investigated the symbolic reasoning abilities of LLMs, focusing on formal language tasks and structured reasoning evaluation.
- Co-authored two papers on LLM reasoning submitted to ACL ARR March 2026.
- Research Collaboration, Apurba-NSU R&D Lab …………………………………. North South University
Focus: Model compression techniques (knowledge distillation) for large language models (LLMs)
Supervisors: Dr. Nabeel Mohammad, Dr. Shafin Rahman
Nov. 2024 – Jun. 2025- Conducted research on model compression techniques (knowledge distillation) for large language models (LLMs).
- Paper accepted to EMNLP 2025 Industry Track as a joint first author, proposing a vision-language reasoning pipeline that maintains competitive performance with significantly reduced computational cost.
Research Assistant (Part-time) — Natural Language Processing (NLP), North South University
Dhaka, Bangladesh · Dec 2023 – Feb 2024
Supervisor: Dr. Mohammad Rashedur Rahman
- Built a context-aware data-cleaning pipeline for Bengali NLP, optimizing for transformer models.
- Achieved up to +4% accuracy on three Bengali benchmarks by preserving sentence-level context and reducing noise; published in SN Computer Science (Springer Nature, IF 4.34).
🧑‍🏫 Teaching Experience
Lab Instructor — Dept. of Electrical & Computer Engineering (ECE), North South University
Feb 2025 – Present
Courses: CSE 225L Data Structures & Algorithms Lab, CSE 115L Programming Language I Lab
- Conduct coding-focused lab lectures; demonstrate DS&A concepts with hands-on exercises.
- Lab Instructor Coordinator (CSE225L): updated manuals, optimized problem sets, standardized lab operations.
- Proctor coding assessments; debug code; evaluate algorithmic efficiency; grade assignments, viva, and lab exams.
Teaching Assistant — Dept. of Electrical & Computer Engineering (ECE), North South University
Feb 2023 – Jan 2025
Courses: CSE 225 Data Structures & Algorithms, CSE 215 Programming Language II
- Mentored 160+ undergraduates per semester in DS&A and OOP.
- Led tutorial sessions and provided targeted support on core topics.
- Graded assignments with detailed, formative feedback.
🏆 Competitive Programming Experience
Programming Contestant — NSU Problem Solvers
Jun 2021 – Feb 2023
Platforms: ICPC, Codeforces, AtCoder, Meta Hacker Cup, Vjudge
- Regular participant in ICPC Asia Dhaka Regional Online Preliminary, received Honorable Mention.
- Solved ~1000 problems across Codeforces, CodeChef, AtCoder, and LightOJ.
- Mentored peers in Data Structures & Algorithms.