I build AI systems for scientific and knowledge-intensive work.

My background combines 14+ years of battery-materials research at Argonne National Laboratory, production data science, and recent AI/ML engineering work on agentic workflows, RAG, evaluation, observability, and MLOps.

Portfolio · LinkedIn · Google Scholar

• Scientific agents for materials discovery and redox-flow-battery formulation
• RAG systems with retrieval evaluation, citation quality, and business-rule aware routing
• Agent harnessing: bounded automation, tool use, QA gates, traces, and repeatable workflows
• MLOps and deployment readiness across Azure AI Search, Azure Container Apps, Docker, observability, and smoke testing
• Turning deep scientific domain knowledge into practical AI workflows

Structured personal research and AI portfolio built with Astro and data-driven JSON records.

It connects publications, patents, talks, collaborators, AI projects, and career narrative into one searchable public profile.

Live: sciencesloop.com

In-progress vertical scientific agent for redox-flow-battery formulation.

The goal is to translate flow-battery chemistry knowledge into evidence-grounded candidate retrieval, formulation reasoning, validation checks, and next-step planning.

Azure-deployed RAG and assistant workflows for energy and enterprise settings.

Focus areas include Azure OpenAI, Azure AI Search, agent workflow design, retrieval quality, and deployment readiness.

Industrial data science work across large operational datasets, model monitoring, SHAP-based explanation, LightGBM modeling, GLM/risk segmentation, ETL validation, and production-readiness checks.

• Former Argonne scientist in battery materials, redox-active molecules, Li-ion electrolyte additives, silicon binders, and redox-flow batteries
• Contributor to autonomous discovery, high-throughput experimentation, robotic chemistry, and ML-assisted electrolyte design collaborations
• 100+ publications, 23 granted U.S. patents, 4,700+ citations, and R&D 100 Award-recognized battery technologies

I like systems that make hard work repeatable:

• clear interfaces before implementation
• small tools that remove recurring friction
• evaluation before claims
• logs, traces, and artifacts over memory
• practical automation with explicit safety boundaries

The long-term direction is simple: apply AI agents and production ML practice to accelerate scientific discovery.