Clayton Strauch

Pumas-AI March 2025 – Present

Intern

• Developed a retrieval-augmented generation (RAG) pipeline using OpenAI LLMs to generate support ticket responses for a pharmacometric modeling software package.
• Reduced manual workload and improved response times for support team by auto-drafting replies using contextual data from prior support threads, documentation, and tutorials.
• Integrated email service APIs for real-time input/output and leveraged Pinecone for semantic search and vector-based document retrieval.
• Tuned prompt templates and retrieval strategies through iterative evaluation to ensure clarity, relevance, and accuracy.

Pacific Northwest National Laboratory (PNNL) May 2021 – November 2024

Research Associate November 2022 – November 2024

• Balanced multiple technical projects while independently delivering timely results and deliverables.
• Co-authored project proposals, technical work plans, sponsor presentations, and peer-reviewed research publications.
• Engineered on an NLP analytics tool powered by Databricks and PySpark to process large-scale, distributed textual datasets.
• Designed and executed unsupervised clustering experiments to rank document similarity and improve retrieval accuracy.
• Built a custom NLP pipeline leveraging domain-specific Transformer models to compute text relevancy scores.
• Developed Python scripts to mine, clean, and standardize biomedical literature from multiple scholarly databases.
• Integrated mining scripts into an end-to-end Python package that also retrieves associated omics data via web scraping.
• Wrote custom Torch DataLoader classes for Transformer text models. Retrieved activation layers from models.
• Enhanced a suite of R packages for mass-spectrometry data management and visualization.
• Implemented a scoring algorithm to recommend appropriate machine learning models to users based on their data and goals.

Intern May 2021 – November 2022
National Security Internship Program

• Developed a retrieval-augmented generation (RAG) pipeline using OpenAI LLMs to generate support ticket responses for a pharmacometric modeling software package.
• Reduced manual workload and improved response times for support team by auto-drafting replies using contextual data from prior support threads, documentation, and tutorials.
• Integrated email service APIs for real-time input/output and leveraged Pinecone for semantic search and vector-based document retrieval.
• Tuned prompt templates and retrieval strategies through iterative evaluation to ensure clarity, relevance, and accuracy.

University of Southern California - SPEC Lab May 2020 – September 2020

Research Assistant
NSF-Funded Research Experience for Undergrads (REU)

• Used R to obtain, clean, format, wrangle, and visualize raw data for merging with a larger database.
• Led group sessions teaching and troubleshooting R for other research assistants.