Building and optimizing large-scale ML and LLM systems to personalize audiobook experiences for millions of listeners.

Key work includes:

• Developing and maintaining production-grade machine learning and semantic search systems that drive personalized recommendations.

• Creating reusable data and API components to improve scalability, reliability, and development velocity across multiple ML pipelines.

• Integrating large language models into core workflows to enhance audiobook metadata, discovery, and summarization.

• Applying prompt engineering and evaluation techniques to improve generative output quality and alignment with listener preferences.

• Optimizing distributed workflows and caching layers to reduce latency and improve compute efficiency.

• Leading team efforts on AWS security compliance, access control audits, and OnCall operational excellence initiatives.

• Supported the Data Science Engineering team in developing internal tools to accelerate ML experimentation and model evaluation.

• Built interactive React and D3.js dashboards that visualized model metrics and surfaced insights for data scientists.

• Collaborated cross-functionally to improve the interpretability and usability of recommendation and NLP models.

• Assisted instruction in introductory Computer Science and Data Science courses for undergraduate students.

• Guided labs and provided one-on-one support on programming fundamentals and algorithmic problem-solving in Python and Java.

• Helped evaluate assignments and coursework for classes of 100 + students per term.

• Developed control and analysis software for the Proteus focused-ultrasound research platform used in the NeuroFUS laboratory.

• Designed real-time acquisition and visualization tools in Python for experiment monitoring and data collection.

• Engineered a custom USB communication module to replace PyVISA, eliminating licensing costs and increasing portability.

• Created PyQt interfaces that simplified transducer configuration and automated experimental workflows.