I'm Sepehr Rezaee, a Bachelor’s student in Computer Science at Shahid Beheshti University (2021–2025), Tehran. My current research and professional interests center around:

• Deep Learning (especially Computer Vision, Generative AI, AI Safety, and AI Agents)
• Physics-Informed Neural Networks (PINNs) and differential equation modeling
• Secure and robust machine learning, including backdoor and trojan detection

• AI Specialist (LLMs) @ PropTy Global (2024–Present)

  • Developed an AI-powered chatbot and home-finder system with open-source LLMs (GPT-Neo, GPT-J, LLaMA) and Retrieval-Augmented Generation.
  • Engineered a scalable backend using FastAPI, Docker, Kubernetes, PostgreSQL, and MongoDB.
  • Achieved significant improvements in chatbot accuracy, scalability, and real-time performance.

• Research Assistant @ Robust & Interpretable Machine Learning Lab, Sharif University of Technology (2024–2025)

  • Contributed to multiple papers (3+ submitted to NeurIPS 2024) on model reliability and security.
  • Built robust ML pipelines to increase model reliability under adversarial conditions.
  • Presented research findings at international conferences.

• Research Assistant @ AI & Scientific Computing Lab, Shahid Beheshti University (2023–2025)

  • Co-authored several papers (under review & accepted) on disease progression modeling, Fokker-Planck equations, and Alzheimer’s detection using MRI.
  • Leveraged PINNs for integrating physical laws into neural networks, improving scientific computing accuracy.

• Deep Learning & Neuroscience Intern @ Institute for Research in Fundamental Sciences (IPM) (2023–2024)

  • Performed comprehensive M/EEG data analysis to decode neural signals using advanced deep learning.
  • Streamlined preprocessing and artifact removal for efficient, high-impact neuroscience research.

• Scanning Trojaned Models Using Out-of-Distribution Samples (Accepted to NeurIPS 2024)
• Comparison of Pre-training and Classification Models for Early Detection of Alzheimer’s Disease Using MRI (Accepted in I4C 2023)
• Multiple ICCV 2025 submissions on data-free backdoor detection, robust non-convex optimization, and novelty detection under style shifts.

Full details and additional papers can be found on my Google Scholar and CV.

• Head Teaching Assistant for Advanced Programming and Data Mining & Analysis at Shahid Beheshti University.
• Teaching Assistant for Basic Programming.
• Mentored projects on AI applications in petrochemical, water, and electricity industries.

• Best Ideator Award at the 7th National Young Scientists Festival (2023) for designing an AI-based assistant for early Alzheimer’s disease detection.
• Placed 352nd among ~150,000 participants in the national entrance exam (2020).

• Email: [email protected]
• GitHub: SepehrRezaee
• LinkedIn: linkedin.com/in/sepehr-rezaee/
• Website: sepehrrezaee.github.io