About Me

I'm currently a 2nd-year PhD Student in computer science at the University of Maryland, advised by Prof. Furong Huang. I have experience doing research in deep learning, large language models (LLMs), agentic AI, security/privacy, and blockchains. My current research focuses on the intersection of security/privacy and machine learning, particularly in the context of large language models and agentic AI. I am always looking for new opportunities to collaborate on exciting research projects, so feel free to reach out if you have any ideas!

Publications
(☆ denotes equal contribution)

Skills

Programming

Python (Full Proficiency), Java (Full Proficiency), C/C++ (Full Proficiency), Rust (Intermediate Proficiency), Go (Intermediate Proficiency), Javascript (Intermediate Proficiency), Android (Java), Solidity (Intermediate Proficiency), Shell/Bash Scripting

Technologies

Docker, Virtualization, Parallel & Multi-Threaded Computing, Linux (RHEL/Fedora, Debian), Database Design, SQL, NoSQL, Kernel Development, CUDA, Git, SLURM, ROS (Robot Operating System), LaTeX

Deep Learning

PyTorch, TensorFlow, LLMs, Transformer Models, Model Quantization, Model Pruning, LLM Efficiency, Membership Inference Attacks, Security and trustworthiness, Microsoft Autogen, Agentic Frameworks

Networking & Security

Blockchains, Smart Contracts, zk-SNARKs, zk-STARKs, Elliptic Curves, Wireshark, Metasploit, Nmap, OWASP/MITRE Methodologies Analysis, Kali Linux

Soft Skills

Communication (Full Proficiency), Teamwork (Full Proficiency), Leadership (High-Intermediate Proficiency), Critical-Thinking (Full Proficiency), Problem-Solving (Full Proficiency), Adaptability (High-Intermediate Proficiency)

Research Experience

Graduate Researcher, University of Maryland

Aug 2023 - Present
Under Supervision of Prof. Furong Huang

  • Submitted the AegisLLM paper (equal-contribution first-author) to the ICML 2025 Conference. This paper is pending a funding acknowledgment after which it will be Arxiv'ed (very soon). We expect the reviews to be very positive based on the strong results presented.

  • Collaborating with Dr. Udari Madhushani Sehwag from JPMorgan AI Research and Stanford University on the security of AI agents under certain privilege escalation scenarios, which is crucial for ensured security in agentic environments. This work (equal-contribution first-author) is almost done and is to be submitted to NeurIPS 2025. We use Microsoft Autogen as our agentic framework.

  • Based on the observation that current PTQ quantization approaches (s.a. GPTQ, AWQ) dramatically lower the performance of large language models for test-time scaling (TTS), working on a new post-training quantization approach specifically optimizing for TTS performance (BoN, beam search, weighted majority voting). This work is targeted for NeurIPS 2025 and is expected to be ready by late-Mar to mid-Apr 2025.

  • Working on a novel problem for on-the-fly single-shot fine-tuning of Large Language Models (LLMs) using diffusion-based approaches.

Research Intern, TU Graz and Universität Klagenfurt

June 2022 - Aug 2023
Under Supervision of Prof. Martin Gebser and Dr. Pierre Tassel

Developed a supervised approach to solving the RCPSP by using imitation learning for learning from CP schedules on smaller instances and generalizing to larger ones. This work is in continuation of Reinforcement Learning of Dispatching Strategies for Large-Scale Industrial Scheduling.

Research Assistant, Data Analysis Lab, University of Tehran

June 2021 - Oct 2022
Under Supervision of Prof. Behnam Bahrak

As part of the Crystalline project, designed and implemented RPoA, a flexible service-based alternative to the currently in-use consensus protocols, as mainly inspired by the drawbacks of PoA. Also, suggested and studied a novel node rewarding policy for full nodes in blockchain networks.

Research Intern, EMC Lab, EPFL

Nov 2021 - May 2022
Under Supervision of Dr. Hamidreza Karami

Studied acoustic source localization using single sensors by modeling using transformer models.

Research Intern, Robotics Lab, Arak University

June 2021 - Nov 2021
Under Supervision of Prof. Mohsen Rahmani

Studied a LiDAR-based context detection and planning problem for mapless autonomous driving using CNNs and UNets.