Owned an embedded AI documentation assistant for Mission Control, Disney's ad campaign platform, from early design spikes through production rollout.

• Owned an embedded AI documentation assistant for Mission Control (ad campaign platform) from spikes to production.
• Authored design spikes on the agent framework, AWS Bedrock model selection, and the streaming wire contract.
• Built a Python/FastAPI service running a Strands agent on Bedrock, answering questions via RAG over a docs knowledge base.
• Integrated a React/TypeScript widget, a GraphQL/WebSocket gateway, and cross-account IAM; deployed to dev, QA, and prod on Kubernetes.

Research on dataset integrity and real-time eye tracking for classroom learning experiments.

• Quantified the impact of mislabeled training data on model performance, producing findings that resulted in a publication.
• Developed eye-tracking software that streams and processes gaze data for live analysis, enabling new intervention experiments.
• Architected asynchronous processing and multi-device concurrency to maintain reliability under simultaneous data streams.

Deep learning for forecasting heart failure events from wearable and clinical data, submitted as co-author to JAMIA OPEN.

• Trained deep and ensemble learning models (LSTM, XGBoost) in PyTorch to forecast heart failure events from Fitbit and clinical data.
• Conducted time-series analysis via statistical testing and visualization to support a co-authored submission to JAMIA OPEN.
• Engineered Python data pipelines (pandas, NumPy) to clean and align irregular wearable and clinical records for ML training.
• Reviewed transformer-based (BERT) and wearable data literature to guide model design and support manuscript writing.

Teaching assistant across core computer science courses, supporting 200+ students.

• Leads weekly office hours (3–4 hrs) for multiple courses including Algorithms and Programming Languages, supporting 200+ students.
• Partners with professors to grade hundreds of assignments and exams, providing feedback and clarifying challenging concepts.