Robust AI
Designing attacks and defenses for image models, diffusion pipelines, and multimodal RL agents.
AI/ML researcher - PhD candidate - Portland, Oregon
Engineering Robust ML for Vision, Language, and Agents.
I am Shayan Jalalipour, a Computer Science PhD candidate at Portland State University working across machine learning, reinforcement learning, computer vision, generative models, and AI code reliability.
AI and ML roles
Research depth. Practical delivery.
Generative systems
Working with diffusion models, transformers, LLMs, multimodal models, and RL environments.
Computer vision
Using object detection, GIS, LiDAR-derived data, and spatial analysis for applied ML problems.
ML engineering
Building with Python, PyTorch, CUDA, TensorFlow, Hugging Face, Docker, Kubernetes, SQL, AWS, and GCP.
Current work
AI Research, Data Science, and Teaching
Jun 2022 - Present
Machine Learning Research Assistant - Portland State University
NSF-funded research in computer vision, generative models, adversarial robustness, and reinforcement learning.
Oct 2025 - Present
Handshake MOVE Fellow - Handshake
Reviewing LLM-generated ML code for scientific validity, edge cases, and reproducibility risks.
Sep 2021 - Present
Teaching Assistant - Portland State University
Supporting courses in Reinforcement Learning, Virtual Reality, and Natural Language Processing.
Jun 2019 - Sep 2019
Data Scientist - Vacasa
Built data analysis tools and geospatial pipelines for operational data.
Publications
Publications in Adversarial ML, Multimodal RL, Diffusion, and Computer Vision.
2026
Latest publication
Understanding Adversarial Vulnerabilities and Emergent Patterns in Multimodal RL
Studies adversarial vulnerabilities and emergent behavior in multimodal RL agents.
2026 International Conference on Semantic Computing (ICSC)
Read on IEEE Xplore2025
OSA-Diff: An Origin Sampling Based Adversarial Attack Using Diffusion Models
Introduces a diffusion-based adversarial attack with faster convergence and lower training overhead.
2025 19th International Conference on Semantic Computing (ICSC)
Read on IEEE Xplore2023
Noisy-Defense Variational Auto-Encoder (ND-VAE): An Adversarial Defense Framework to Eliminate Adversarial Attacks
Presents a VAE-based defense for removing adversarial perturbations from images.
2023 Fifth International Conference on Transdisciplinary AI (TransAI)
Read on IEEE Xplore2023
Deep Learning-Based Spatial Detection of Drainage Structures Using Advanced Object Detection Methods
Evaluates Faster R-CNN, DINO, DETR:DINO, and YOLOv5 for drainage-structure detection.
2023 Fifth International Conference on Transdisciplinary AI (TransAI)
Read on IEEE Xplore