(2at) Accelerated Materials Design and Discovery Using Self-Driving Laboratories.

Advances in high-throughput experiments and widely available data management and analysis tools resulted in the development of self-driving laboratories (SDL) that aim to accelerate the design and discovery of key materials by several orders of magnitudes. To fully unlock the potential of SDL, we need a better way of processing experimental data, and workflows that are focused on knowledge extraction. My research interests lie at the intersection of computers and experimental material science going beyond traditional tools to explain the observed phenomenon to empower and accelerate the data analysis, workflows, design, and discovery of materials to address key societal and economic challenges. My vision to achieve this is by developing an interdisciplinary research and teaching program focused on empowering students with modern usage of data-driven computational methods focusing on open-source dissemination of information.

Prior work:

In my Ph.D. work, I investigated the role of data representations as a scientific modeling tool in realizing the goal of autonomous design and discovery using self-driving laboratories. Although spread across different applications, materials, and data sources, a common theme for Ph.D. work is the role of data representations to encode, process, and evaluate data coming from scientific instruments so that modern data-driven methods can be used to build autonomous workflows that can accelerate synthesis, design, and discovery of materials. In my postdoctoral work, I combined my background in developing closed-loop frameworks and mathematics of data representation to build a real-world scientific experimental workflow for retrosynthesis, knowledge extraction, and phase mapping of nanoscale colloidal and polymer materials. The function space-based data analysis for experimental data opened up a new avenue and is significantly better at modeling complex phase and morphological transitions within the scope of autonomous experimental workflows.

As an assistant professor, my research vision focuses on pushing the boundaries of data-driven models and data representations as a modeling tool for in-line experimental design and discovery. Our research group’s philosophy will revolve around the theme of ‘computations for experiments’ building tools and collaboration opportunities for experimentalists to use computational tools for data analysis, management, and workflows. In particular, the research group will focus on accelerating knowledge extraction rather than improving upon the predictive performances of existing, purely computational models. To this end, we will combine our expertise in data-driven and physics-based modeling to provide faster, differentiable simulation tools that work asynchronously with experimental workflows to retrosynthesis using global back-box optimization, extraction of design rules using differentiable models of experiments, and construction of stochastic models of phase diagrams. In this poster, I will highlight these directions and discuss specific methods that are suited to solve them.

Teaching Interests:

My personal experiences made me understand the importance and influence I can have on others as a teacher and mentor as I progressed through my early career. I believe we should teach students not just how to solve a scientific problem but rather foster an institution for creatively and responsibly solving global problems and efficiently communicating them to various stakeholders.

My teaching interests broadly cover computational and numerical methods in material science and chemical engineering. I am particularly well suited to teach courses with data science, numerical computing, and modeling components included in them. Recently, I designed and delivered a new course of design experiments for material scientists and chemical engineering graduate students, and have experience in mentoring undergraduate and graduate students.

Selected CV highlights:

• Invited talks: ACS Fall Meeting 2023, IEEE Rochester Chapter
• Awards: UW Data Science Postdoctoral Fellow and Conference Travel award
• Teaching: Primary instructor (CHEM E 599), Project mentor DIRECT (CHEME / CHEM / MSE 547) University of Washington

Selected publications:

[1] Vaddi, K., Liu, H., Pokuri, B. S. S., Ganapathysubramanian, B., and Wodo, O. (2023) “Construction and high throughput exploration of phase diagrams of multi-component organic blends” Computational Materials Science 216 (2023) : 111829.

[2] Vaddi, Kiran, Karen Li, and Lilo D. Pozzo. “Metric geometry tools for automatic structure phase map generation.” ChemrXiv, pre-print (2022)

[3] Vaddi, Kiran, Huat Thart Chiang, Lilo D. Pozzo “Autonomous retrosynthesis of gold nanoparticles via spectral shape matching” Digital Discovery (2022)

[4] Lachowski, K. J., Vaddi, K., Naser, N. Y., Baneyx, F., Pozzo, L. D. “Multivariate Analysis of Peptide-Driven Nucleation and Growth of Au Nanoparticles” Digital Discovery (2022)

[5] Vaddi, Kiran, Olga Wodo “Active knowledge extraction from cyclic voltammetry” Energies 15.13 (2022) : 4575

[6] Vaddi, Kiran, and Olga Wodo. “Metric Learning for High-Throughput Combinatorial Data Sets.” ACS Combinatorial Science 21.11 (2019) : 726-735