Joseph Bruechner

software developer

Python, scikit-learn, Ridge Regression

Hospital LOS Optimization

Machine learning solution predicting hospital length of stay with 86% accuracy (R² 0.86, RMSE 2.57 days). Identifies key factors like emergency admissions (+3 days), COPD (+2.9 days), and weekend effects. Includes practical implementation roadmap with expected savings of $2.5-3.8M annually for mid-sized hospitals through targeted interventions and resource optimization.

Python, pandas, numpy, OOP

ML Evaluation Framework

Flexible, object-oriented framework for evaluating ML models across multiple dimensions including helpfulness, safety, and performance efficiency. Features modular architecture with BaseEvaluator, ScoringEvaluator, and CompositeEvaluator classes. Enables trade-off analysis revealing model compromises (e.g., 76.67% both helpful and safe) with scalable parallel execution for large-scale assessments.

C++, Threading, Real-time Systems

Rocket Test Automation

Dual C++ system combining test stand automation with rocket simulation. TestStandAutomationTool manages parallel test execution with safety monitoring and real-time data processing. RocketTestSimulator generates realistic test data with configurable parameters for thrust, pressure, and temperature profiles. Both feature robust error handling and comprehensive logging systems.

GCP, BigQuery, Python, Looker

SaaS Analytics Pipeline

Complete end-to-end data pipeline for SaaS business analytics built on Google Cloud Platform. Integrates BigQuery for data warehousing, Python for ETL processing, and Looker for interactive dashboards. Processes customer lifecycle metrics, revenue analytics, and churn prediction with automated data quality checks and real-time monitoring capabilities.

Python, Random Forest, Time Series

Bitcoin Price Prediction

Random Forest model predicting Bitcoin prices with ~$200 deviation. Processes 6M+ records of 1-minute interval data from 2012-present, resampled to 15-minute intervals. Features technical indicators, OHLCV data, and time-based features. Achieved R² of 0.9755 with real-time prediction capability and comprehensive backtesting framework.

Python, React, TypeScript, NLP

Atticus AI

Full-stack production legal AI platform helping lawyers and legal professionals accelerate workflows without compromising quality. Custom Python API backend with advanced NLP processing, AI powered RAG pipeline, TypeScript/React frontend with modern UI/UX design. Features document analysis, case research automation, and intelligent legal drafting tools with enterprise-grade security and compliance.