My expertise lies in building AI-powered experimental platforms that accelerate scientific R&D, with a current focus on materials design and characterization. With a proven track record of collaborating closely with academic and industry partners, I excel at translating complex domain-specific challenges into efficient and reusable AI-first infrastructure. My technical contributions include developing autonomous microscopy platforms, creating active hypothesis learning framework for automated experiments, and authoring major open-source tools - such as AtomAI, GPax, and SciLink - that integrate machine learning and agentic AI into physical experiments. Additionally, I pioneered the "Jupyter paper" concept to enhance research transparency and reproducibility. Ultimately, my goal is to empower human-AI collaboration for accelerating scientific innovation and real-world applications.
- Unknown Knowns, Bayesian Inference, and structured Gaussian Processes
- Deep Learning Meets Gaussian Process: How Deep Kernel Learning Enables Autonomous Microscopy
- Gaussian Process: First Step Towards Active Learning in Physics
- Mastering the shifts with variational autoencoders
- "Dynamic STEM-EELS for Single-Atom and Defect Measurement During Electron Beam Transformations." Science Advances (2024). Contribution: Developed a deep learning-based rapid object detection and action system (RODAS) and oversaw its implementation on a multi-million-dollar electron microscope.
- "Experimental Discovery of Structure-Property Relationships in Ferroelectric Materials via Active Learning." Nature Machine Intelligence (2022). Contribution: Developed an automated workflow for active learning of the relationship between local structures and physical properties in multi-modal experiments.
- "From Atomically Resolved Imaging to Generative and Causal Models." Nature Physics (2022). Contribution: Introduced AI-driven extraction of domain-specific information from microscopy data for building generative models over a broader parameter space and exploring causal mechanisms underpinning functionalities.
- "Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries." Advanced Materials (2022). Contribution: Developed an active hypothesis learning approach based on co-navigation of the hypothesis and experimental spaces in automated experiments, allowing physics discovery via active learning of competing hypotheses.
- See the full list here
- Ziatdinov, Maxim A., et al. "Science-driven automated experiments." U.S. Patent No. 11,982,684. 14 May 2024.