AI Systems2025· Internship
AI Operational Platform
Kaya AI
Leading development of two interconnected systems: Magellan, a real-time editable dashboarding platform for construction and supply chain operations; and Amber, an AI analyst layer that enables natural language querying over enterprise operational data. Both systems are deployed with live enterprise customers.
→ impact
Translating fragmented construction workflows into structured, queryable data products used by non-technical operators daily.
ReactTypeScriptPostgreSQLLLMsRAGNL→SQL
Data Infrastructure2024· Internship
Snowflake Usage Audit Dashboard
NetJets
Built a daily-refresh audit system to surface table usage patterns, never-queried assets, and compute costs across the entire Snowflake estate at one of the world's largest private aviation companies. Gave analysts and leadership real transparency into data health for the first time.
→ impact
Identified significant compute waste from abandoned tables and unused queries. Enabled data team to prioritize cleanup and reduce Snowflake spend.
SnowflakeSQLPythonTableau
Applied ML2025· Internship
Manufacturing ML Feature Pipeline
Kalypso × Rockwell Automation
Designed and implemented a feature engineering pipeline to support model development for real operational decision-making in a manufacturing environment. Built to operate within actual industrial automation constraints — not a sandbox dataset.
→ impact
Pipeline supported model evaluation across multiple manufacturing use cases in Rockwell Automation's production environment.
PythonFeature EngineeringScikit-learnIndustrial ML
Analytics2023· Internship
Ticket Traffic Dashboard
Ohio State Department of Athletics
Created a ticket-traffic dashboard to track demand trends and performance patterns across athletic events. Designed specifically for non-technical stakeholders who needed to monitor activity and make decisions without analyst support.
→ impact
First dedicated analytics tool for the athletics ticketing operations team. Enabled self-serve monitoring and faster response to demand signals.
SQLDashboardingAnalyticsStakeholder design
Operations Research2024· Course
NFL Schedule Optimization Model
Systems Modeling & Optimization
Formulated and solved a schedule optimization model for NFL game planning. Encoded real constraints — travel distance, rest days, primetime slot allocation, and divisional game balance — then analyzed the tradeoffs that influence projected win probability.
→ impact
Demonstrated that naive scheduling leaves 8–12% win probability on the table relative to optimized schedules under the same constraints.
GurobiLinear ProgrammingPythonConstraint modeling
Deep Learning2024· Course
Facial Emotion Recognition
Neural Networks — OSU CSE
Trained and compared CNN architectures for facial emotion recognition. The project focused on rigorous evaluation: confusion matrices, per-class error rates, and tradeoff analysis between model complexity and performance under compute constraints.
→ impact
Established that transfer learning with lightweight fine-tuning outperformed from-scratch training by ~14% accuracy at fraction of the training cost.
PyTorchCNNsTransfer learningError analysis