Summer 2026 Technology Internship Program

Developed the organization's most accurate winter peak demand forecast by implementing an XGBoost model with tailored meteorological features and data center load-growth analysis, reducing MAPE by 47%

Reduced data retrieval time by ~97% by developing a Python CLI tool leveraging the AVEVA PI Web API, replacing a manual Excel-based ingestion process with automated shell scripts for monthly analyses.

Engineered a real-time AVEVA PI Vision dashboard measuring data center load swings' impact on Area Control Error (ACE) deviations, giving management the evidence needed to pursue new operational regulations.

Enabled identification of undiagnosed patients and risk stratification by building a BERT-based model that encodes longitudinal claims data into dense embeddings, supporting patient journey clustering across large populations.

Strengthened risk assessment for an anti-seizure medication by developing a survival analysis model with left-truncation and right-censoring corrections, benchmarking internal outcomes against real-world evidence.

Validated drug usage patterns for a rare pediatric epilepsy disorder by running subpopulation analyses on Komodo claims data in Snowflake, surfacing prescribing trends that informed commercial strategy.

Enabled faster executive decision-making by building a LangChain-based AI analytics agent that translates natural language questions into business metrics and automatically generates dashboards.

Increased agent reliability by curating a human-validated question-to-SQL dataset and implementing Milvus semantic retrieval to surface similar queries, increasing accuracy by 16%.

Scaled client-to-competitor product matching across thousands of SKUs by migrating to batched embeddings via the OpenAI Batch API, cutting embedding costs by 50%.