Software engineer · AI agents & ML
Patrick Beal
I build models, and I also build the systems that run them.
I'm a software engineer who builds AI agent systems, with a background in machine learning, data science, and economics.
Two lanes
Systems
The apps, pipelines, and AWS infrastructure I have built, mostly as a founding engineer.
Models
Bayesian modeling, causal inference, and forecasting, mostly from my research and team projects at BYU.
About

I like to work on a problem from the analysis all the way through to software that someone can use, which usually means doing the modeling and also building the system that runs it. My background is in optimization, machine learning, Bayesian methods, and econometrics, and over the past year I have added AWS, Terraform, and LLM applications on top of that.
Recent work
all work →| role | status | dates | highlight |
|---|---|---|---|
| Teradata Software Engineer | current | Sep 2026 – Present | — |
| TechForce Advisors Founding Engineer | Aug 2025 – Sep 2026 | $109 to $0 peak monthly NAT Gateway cost | |
| Winning by Design GTM Strategy & AI Intern | May 2025 – Sep 2025 | 6–8 wk → 5–15 min diagnostic turnaround | |
| BYU Office of the President Research Assistant | Aug 2024 – Dec 2025 | 6 RAs team |
Selected projects
all projects →Strip under each project:highlightedstage covered,dimmednot covered
- Agentic Company Research Systemsystem · 2026
A production system that turns a visitor's email into a personalized company report, then checks the report twice for unsupported claims before sending it.
result A production system that double-checks every report before sending it, with per-report cost tracking.
- data (covered)
- model (covered)
- eval (covered)
- serve (covered)
- monitor (covered)
- Grounded Document Extractionsystem · 2026
A pipeline that extracts financial data from real estate broker documents, where every value has to cite the page and text it came from.
result I built the parser, the routing, the citation checks and the evaluation, and the agent loop has not yet accepted values on real documents. The project is stalled.
- data (covered)
- model (covered)
- eval (covered)
- serve (not covered)
- monitor (not covered)
- Shared AWS Test Platformsystem · 2026
One Terraform-managed test environment shared by three apps: a single load balancer routes by hostname, one Aurora cluster serves them all, and a NAT instance replaces the NAT Gateways.
result Three apps share one test environment, and replacing NAT Gateways with NAT instances cut that line from as much as $109 a month to a few dollars.
- data (not covered)
- model (not covered)
- eval (not covered)
- serve (covered)
- monitor (covered)
- Biomarker Detection in PPG Signalsmodel · 2024
Research on extracting biomarkers from photoplethysmogram (PPG) signals, including peak detection, signal-quality metrics, and a Bayesian mixture-of-experts ensemble.
result My signal-resampling module and its test suite were merged into the research group's shared codebase and adopted lab-wide.
- data (covered)
- model (covered)
- eval (covered)
- serve (not covered)
- monitor (not covered)
A three-person team project modeling housing-affordability trends across U.S. states from 2000 to 2020, using time-series methods and Bayesian hierarchical modeling on Zillow price and CPS demographic data.
- data (covered)
- model (covered)
- eval (covered)
- serve (not covered)
- monitor (not covered)