Summary
Physics PhD turned data scientist, with five years building and deploying end-to-end ML systems that move the needle. I spent a decade in experimental and computational physics before moving into industry. Since then I’ve designed deep learning architectures, LLM pipelines, and automated ML workflows for patent analytics and financial data — work that has directly driven hundreds of thousands of dollars in revenue. I’m drawn to problems where the modeling is genuinely hard, and I’ll happily go as deep as the problem requires to get it right.
Work Experience
Staff Data Scientist — Moat Metrics, Spokane, WA (Jun 2024 – Mar 2026)
- Built an end-to-end patent valuation pipeline: trained a PyTorch MLP to predict price-to-earnings from patent embeddings aggregated at the portfolio level, with polynomial bias correction to calibrate output distributions against actual PE values; deployed on AWS Batch with weekly automated retraining and daily inference runs across thousands of entities using Polars-based data retrieval from Snowflake
- Engineered LLM pipelines for large-scale patent text processing, including few-shot and chained prompting, structured output extraction, and local LLM inference via vLLM with KV caching to optimize throughput; deployed batch workloads on GCP
- Created an ETL pipeline for a Snowflake database linking patents to government contracts, including parsing and cleaning of malformed patent IDs and contract numbers from heterogeneous sources, resulting in a $250,000 external contract
- Led a data wrangling and analysis exercise contracted by a high-profile private equity firm, delivering results that generated $200,000 in revenue above the base contract
- Contributed on a project constructing a knowledge graph on 5 million active US patents through batch LLM parsing, local LLM calls, and Boolean satisfiability (SAT) algorithms
Research Data Scientist — Aon PLC, Spokane, WA (Sep 2021 – Jun 2024)
- Developed an Adaptive Clustering System with a human-in-the-loop retraining loop: user-corrected labels backpropagate through a jointly trained deep autoencoder and classifier via masked loss, with convergence-based stopping criteria that halts training when user overrides are satisfied rather than at a fixed epoch count
- Built the supporting training infrastructure: S3-based checkpoint/resume system supporting incremental retraining across sessions, and a coordinate warping system that visually tightens cluster separation in the projected embedding space
- Built an entity alignment system using Elasticsearch blocking and Conditional Random Fields to probabilistically resolve patent entity identity across heterogeneous data sources
- Implemented graph traversal algorithms to reconstruct patent assignment histories across complex ownership chains
Senior Research Associate (Postdoctoral) — Duke University, Durham, NC (Jun 2019 – May 2021)
- Characterized trapped ions via microwave and optical transitions; optimized PID laser stabilization and performed high-sensitivity interferometric vibration analysis of the trapping apparatus
- Mentored 3 graduate students across physics and ECE departments
Senior Research Associate (Postdoctoral) — University of Oregon, Eugene, OR (Jun 2018 – Jun 2019)
- Conceived and modeled a quantum network based on diamond defect centers, characterizing vibrational resonances and band structure via COMSOL Multiphysics and building a scanning confocal microscope for defect measurement
Research Assistant (PhD Candidate) — University of Oregon, Eugene, OR (Jun 2010 – Jun 2018)
- Built Python simulation libraries for light-matter interactions and designed automated data acquisition systems integrating hardware and software for real-time experiment monitoring
- Directed 4 major projects end-to-end (theory → experiment → publication); contributed to 17 peer-reviewed publications and 2 patents
Education
PhD in Physics — University of Oregon, Eugene, OR (June 2018) · GPA: 3.88
Bachelor of Science in Physics — Washington State University, Pullman, WA (May 2009) · GPA: 3.5
Skills
Programming & Tools: Python, SQL, COMSOL Multiphysics, PyTorch, Polars, Numpy, Matplotlib, Scikit-Learn, Statsmodels, DuckDB, Postgres, Snowflake, Elasticsearch, Git, Docker, Virtual Environments and UV, AWS (EC2, S3, Batch), GCP, vLLM, Gitlab CI/CD Pipelines, Terraform (limited)
Data Science & ML: Deep Neural Networks, LLM Prompt Engineering (zero-shot, few-shot, chaining), Vector Search and Embeddings, Approximate Nearest Neighbors (ANN), Signal Processing, Time-series Analysis, Regression, KMeans and HDBSCAN Clustering, KNN Classification, Conditional Random Fields, Human-in-the-Loop, Active Learning, Joint Optimization, Metric Learning, Manifold Learning, Bayesian Inference
Soft Skills: Scientific writing (17 peer-reviewed publications with over 2,000 citations), cross-disciplinary collaboration, mentorship, technical communication
Awards & Honors
- Marthe E. Smith Memorial Science Scholarship · 2012
- Weiser Sr Teaching Assistant Award · 2016
- The Optical Society Quantum Optical Science & Technology Technical Group Outstanding Poster Presentation Award · 2017