AI for Science · LLM Agents · Scientific Workflow Orchestration
AI-for-Science researcher and ML engineer building LLM agents, retrieval-augmented scientific workflows, and practical AI systems for materials, chemistry, and energy.
Profile
My current work centers on LLM-based multi-agent systems, retrieval-augmented scientific workflows, AI co-scientists, and simulation-centered research automation. I am especially interested in how language models can coordinate scientific reasoning, tool use, evidence retrieval, debate, validation, and experiment planning in materials and chemistry-oriented research.
Research
Primary work
Equal-contribution author and core AI-side contributor to a multi-agent AI co-scientist system for materials inference and discovery. Led direct implementation most substantially in Phase 1: Pre-Experimental Design.
arXiv:2604.13699Research axis
Knowledge-graph-augmented retrieval for chemistry and process-engineering reports, with attention to evidence grounding, TEA/LCA context, domain document handling, and raw retrieval provenance.
Methods
Prompt pipelines, multi-agent discussion, tool-use orchestration, retrieval reliability, structured reasoning, evaluation framing, and research artifact generation.
Outputs
Equal-contribution paper on an LLM-driven multi-agent system for closed-loop materials research, spanning hypothesis formation, simulation planning, debate, expert voting, and validation-oriented workflow design.
Read the arXiv paperChemRAG presentation record for a knowledge-graph-augmented RAG system supporting techno-economic and life-cycle assessment of chemical processes. Listed author; co-first/equal-contribution role within the ChemRAG workstream.
Project-associated contribution as a core ML engineer at Thingspire, connecting drone inspection data, RGB/thermal computer vision, and solar plant operation and maintenance workflows.
Experience
Agentic AI systems for simulation-centered scientific research automation, with direct work on MIND and ChemRAG, scientific workflow design, prompt architecture, retrieval, reasoning, raw evidence inspection, and manuscript/presentation artifacts.
Applied AI and computer vision across solar farm inspection, RGB and thermal imagery, geospatial AI, data collection, model development, GPU environments, Docker, and deployment-oriented system preparation.
MATLAB-based image processing and threshold segmentation for experimental thermal and interfacial transport analysis.
Education
Master of Computer Science, Data Science emphasis · 2023
BS Materials Science and Engineering · CS and CSE minors · 2022
LL.B. candidate · Interests in intellectual property, AI governance, and data protection.
Selected Projects
AI co-scientist system for materials research, accepted to ECML-PKDD 2026 Demonstration Track.
PaperComputer-vision workflows for RGB and thermal drone imagery, panel-level inspection, and O&M support.
S&P 500 and macroeconomic indicator analysis with regression, Random Forest, SVM, and LSTM models.
GitHubComputer-vision assessment of chopstick technique using OpenCV, MediaPipe, TensorFlow, and LSTM.
GitHubNLP summarization project using tokenization, stopword filtering, and extractive scoring.
GitHubGeospatial computer vision prototype for rooftop solar panel detection and capacity estimation.
How I Work
I work best at the boundary where a scientific question has to become a real pipeline: defining the research workflow, choosing the retrieval and agent structure, building the prototype, inspecting failure modes, and turning the result into a paper, demo, or deployable artifact.
Skills
LLMs, RAG, multi-agent systems, prompt engineering, PyTorch, TensorFlow, Scikit-learn, OpenCV, MediaPipe, YOLO-family models, segmentation workflows.
Python, C/C++, Java, SQL, MATLAB, JavaScript, Linux, Docker, CUDA environments, FastAPI, Git/GitHub, AWS.
AI for Science, materials discovery, scientific workflow automation, renewable-energy AI, geospatial computer vision, AI governance.
Contact