I build tools people can run and trust.
My deepest work is in machine learning systems and the infrastructure to run them, including a CatBoost model that predicts used-car prices within ~$1,300 at the median across 243k vehicles, served via a FastAPI endpoint with Docker, Kubernetes, and automated testing. Around that sit data pipelines, geospatial analysis, and client-facing applications, including a live tool that searches 127k parcel records and ships an assumptions log with every run so the client can see how the answer was reached.
Before this I spent several years as a Director of Strategic Planning, making calls off tools whose limits I could not see. That is where I learned to build for the people who inherit the tool, not the person who wrote it.
Explore my projects: Erin Weiss Portfolio
View my résumé: Download Here
Reach me at: [email protected]
Used Car Price Prediction — Model Development
Predicts used-car listing prices within ~$1,300 at the median across 243k vehicles, 29 manufacturers, and 5,600+ model variants. Compared Ridge, CatBoost, and FT-Transformer architectures with reproducible experiment management and versioned artifacts.
Python · CatBoost · scikit-learn · TensorFlow · Keras · pandas · Quarto
Used Car Price Prediction API — Production Deployment
Takes the CatBoost model from Part 1 and ships it as a production API: request validation, fuzzy input matching, multi-stage Docker builds, Kubernetes orchestration with autoscaling, Prometheus + Grafana monitoring, and CI/CD via GitHub Actions.
FastAPI · Docker · Kubernetes · GitHub Actions · Prometheus · Grafana · pytest
Demolition Neighbor Notification — Geospatial Tool (Live App)
Turns a contractor's demolition site list into a complete neighbor notification package. Searches 126,958 St. Louis parcels against 500 ft buffers measured from each site's boundary, returning 1,636 addresses across a 30-site project and flagging 134 parcels whose assessor records contradict themselves for field verification. Ships as both a Streamlit app and a CLI running the identical pipeline.
Python · GeoPandas · Folium · Streamlit · GeoParquet · pytest · ruff
Reinforcement Learning — Agent Optimization
Implements RL algorithms to optimize a mobile robot's warehouse path. Covers exploration vs. exploitation, policy optimization, hyperparameter tuning, and reward shaping.
Python · NumPy · pandas · Matplotlib
Real Estate Analysis — Statistical Modeling in R
Analyzes U.S. housing trends (2016–2022) with a multiple linear regression model explaining ~95% of price variation.
R · tidyverse · ggplot2
See all projects → erin-weiss.github.io
