Cameron Gruich

Machine learning engineer and PhD candidate with interest in building large-scale ML and LLM systems for scientific and industrial applications. Experienced in large-scale data pipelines (100M+ samples), deep learning/RL training, evaluation tooling, and production ML systems. Former Microsoft Research intern.

University of Michigan (Ann Arbor, MI)

• Built and maintained a 140M+ sample molecular dataset and high-throughput data pipeline (LMDB/WebDataset) to enable distributed training of deep learning models for collaborators.
• Built evaluation and benchmarking tooling to compare multiple uncertainty quantification methods for graph neural networks, improving reliability and model selection for materials discovery. Github
• Designed and implemented a feature engineering pipeline that converts molecular structures into image-based representations and trained vision models for property prediction.
• Built a GPU-accelerated image classification pipeline for large-scale microscopy data using CNNs. Github
• Designed automated pipelines to generate, simulate, and evaluate thousands of molecule candidates in parallel on HPC clusters for redox flow battery study.
• Worked with cross-department researchers to run chemical simulations/associated data analysis for alkaline redox flow batteries using density functional theory (DFT).

2025 LLM Hackathon for Applications in Materials Science & Chemistry

• Led a 10-person team in a Hugging Face–hosted materials-science hackathon (3-week sprint) to build RedoxFlow, an agentic LLM system that orchestrates tool use across a multi-step workflow (de novo molecule generation → automated DFT input/script prep → post-run Nernstian redox potential extraction), delivering an end-to-end demo; won Vision Award for innovative project idea. Github

• MIDAS: Rapid, Multiplexed Molecular Profiling for Integrated Host-Pathogen Analysis. Yong Jun Lim, Mohammad Asadi Tokmadesh, Matthew Allen, Cameron Gruich, Jun Hee Choi, Changhoon Kim, Nicole Falkowski, Bryan Goldsmith, Kathleen Stringer, Robert Dickson, Ki Wan Bong, Jouha Min. Nature Communications. 2025.
• Sulfonated Benzo[c]cinnolines for Alkaline Redox-Flow Batteries. Siddharth Singh, Jessica L Tami, Cameron Gruich, Allison J Gatz, Jason Dong, Bichlien Nguyen, Jake Smith, Bryan Goldsmith, Anne McNeil, David Kwabi. ACS Applied Energy Materials. 2025.
• Clarifying Trust of Materials Property Predictions using Neural Networks with Distribution-Specific Uncertainty Quantification. Cameron Gruich, Varun Madhavan, Yixin Wang, & Bryan Goldsmith. Machine Learning: Science and Technology. 2022.