CV
Download CV (PDF)
Education
- Ph.D. in Computer Science, Arizona State University, 2024 – Present
- GPA: 3.67 / 4.0
- Co-advised by Prof. Chitta Baral and Prof. Ruoyu Wang
- Expected completion: Spring 2028
- M.S. in AI Engineering – Information Security, Carnegie Mellon University, 2024
- GPA: 3.7 / 4.0
- Advised by Prof. Amir Barati Farimani and Prof. David Varodayan
- B.A. in Computer Science and Economics, Boston College, 2022
- GPA: 3.65 / 4.0
- Dean’s List (First Honor): 2020–2022
Research Experience
- Graduate Research Assistant, Arizona State University (Aug 2024 – Present)
- Conduct research at the intersection of LLM agents, software security, and AI security, with a focus on building reliable and secure agentic systems
- Develop benchmarking, reasoning, and post-training methods for complex agentic tasks, including open-source software compilation and long-horizon tool use
- Investigate jailbreaks and prompt injection in LLM-based systems, developing data-flow-centric vulnerability analysis methods and post-training defenses
- Graduate Research Assistant, Carnegie Mellon University (May 2023 – May 2024)
- Proposed a new masking pretraining method for Force and Energy-Centric Graph Neural Networks
- Pretrained and finetuned GNNs on water molecule and organic molecule datasets, reducing RMSE by up to 38%
- Research Assistant, Peking University (May 2021 – June 2021)
- Evaluated blockchain protocols and smart contract codes
- Constructed test nets to reproduce attacks targeting smart contracts
- Analyzed the effect of regulations on cryptocurrency transactions using on-chain data
Research Interests
- General: Natural Language Processing and its applications in security and software engineering
- Current Focus: LLM post-training, multi-agent systems, and AI for software security
Professional Experience
- Research Intern, Samsung Research America, Mountain View (May 2026 – Aug 2026)
- Conducted research on indirect prompt injection vulnerabilities in LLM-based agentic systems
- Developed a gym-like environment for scalable training-data sampling and generation
- Trained LLMs using supervised fine-tuning and reinforcement learning methods, including DPO and GRPO
- Achieved state-of-the-art performance on AgentDojo and AgentDyn
- Algorithm Development Intern, Sohu Inc., Beijing (June 2021 – Sept 2021)
- Designed and deployed a real-time feed deduplication system for social media
- Leveraged Kafka, Spark, simhash, and Bloom filters for large-scale data streaming
- Reduced repetitive text exposure by ~90% and duplicate video content by ~30%
Technical Skills
- Programming: Python, C, Scala, Java, R, Swift, JavaScript, HTML, CSS
- ML/DL Frameworks: PyTorch, TensorFlow, PyG, Verl, vLLM
- Data & Distributed Systems: Spark, Kafka, Kubernetes
- Cloud & Databases: MySQL, MongoDB, Hive, HBase; GCP, AWS
- Software Development: React, iOS, Docker, MCP
Publications
One first-authored paper submitted to NDSS 2027.
A. Saeidi*, Z. Zhang*, R. K. Singh, et al. "CAST: Critique-Aware Supervision for Training Reliable Long-Horizon Tool-Calling Agents." EMNLP 2026 Main Conference. (*Co-First Authors)
Z. Zhang, A. Bajaj, D. Handa, et al. "Build-Bench: Benchmarking LLM Agents on Compiling Real-World Open-Source Software." Transactions on Machine Learning Research (TMLR).
Z. Zhang, Z. Li, and A. B. Farimani. "Masked Pretraining Strategy for Neural Potentials." AIP Advances, vol. 14, no. 9, Sep. 2024.
J. Li*, Y. Yuan*, and Z. Zhang*. "Enhancing LLM Factual Accuracy with RAG to Counter Hallucinations: A Case Study on Domain-Specific Queries in Private Knowledge-Bases." arXiv:2403.10446. (*Co-First Author)
D. Handa, Z. Zhang, A. Saeidi, S. Kumbhar, and C. Baral. "When Competency in Reasoning Opens the Door to Vulnerability: Jailbreaking LLMs via Novel Complex Ciphers." NeurIPS 2025 Reliable ML from Unreliable Data Workshop.
Teaching
