AI researcher and software engineer — computer vision, reinforcement learning, and generative models, currently working on formal verification of mechanistic interpretability interventions in transformer models.

Focus: AI Safety & Alignment · Formal Verification · Mechanistic Interpretability

🔭 Currently working on formal verification of interpretability interventions on toy transformer models, targeting a top-tier conference submission.

🎓 Lecturer, Dept. of Computer Science, AIUB · Lead developer, in-department Web Development & Automation team

💼 3+ years of professional software engineering — production backend systems handling millions of events/hour

📄 Published in Frontiers, IEEE, and Springer — computer vision, reinforcement learning, and generative models

• cauli-det — modified YOLOv8 for cauliflower disease detection, 91.1% mAP · paper
• marl-cbis-ddsm — cooperative multi-agent RL for mammogram ROI classification · paper

Python PyTorch TensorFlow Java C++ · HuggingFace Ollama LangChain · Spring Boot Kafka Next.js · Docker Git

🔗 Website · Google Scholar · LinkedIn · sazid.uddin@aiub.edu