Bryce Keeler

Davidson Management Honors Program · Expedition EY Scholarship

• Built ad inventory yield models for a global media & entertainment client — pricing ad slots, forecasting slot revenue, and recommending allocation across competing demand.
• Forecasted component failure across a 2,100+ machine construction fleet from onboard sensor telemetry and daily technician inspections; 0.91 AUC, 96% accuracy over 3 months live.
• Trained and deployed models with PyTorch and TensorFlow on Databricks and AWS, orchestrating retraining and inference through n8n.
• Shipped a production generative AI reporting system on the Anthropic Claude API, eliminating 1,500+ hours of manual work weekly with a mandatory human-review checkpoint.
• Automated XML→JSON transformation pipelines processing 10TB weekly across 25 enterprise integrations; contributed to two apps published on Workday Marketplace.

• Engineered a constraint-solving course scheduling application against real-time enrollment data, later purchased and deployed by two major universities.

• Analyzed large datasets with SQL and Python; built Tableau/Power BI/Alteryx dashboards with automated pipelines, cutting manual processing 60%.

• $10,000 merit scholarship, selected from 4,000+ applicants, in data analytics and machine learning.

• Evaluation engine running 25+ evaluators per agent trace: deterministic checks, LLM-as-judge assessments, and safety scans.
• Complexity-based model routing with fallback chains, reducing client LLM spend 47%; Python and TypeScript SDKs across OpenAI, Anthropic, LangChain, LlamaIndex, CrewAI.

• Multi-strategy system scanning 6,000+ assets; walk-forward backtesting with Sharpe, max drawdown, and win-rate validation across market regimes.

• Random Forest health classification, LSTM soil-moisture forecasting, and an ensemble for watering recommendations over an ESP32 sensor network; OpenCV segmentation cut hardware cost 67%.