Built an end-to-end real-time analytics pipeline ingesting Spotify API data via Kafka, processing streams with PySpark, orchestrating batch jobs with Airflow, and transforming data in BigQuery using dbt, serving trend insights through an interactive D3.js dashboard deployed on GCP.

Built a full-stack React/Flask app integrating multiple LLM APIs (Groq) into a code-execution pipeline, applying React state memoization and performance tuning to handle 300+ queries per period while minimizing render overflow; developed AI-assisted with Claude and Cursor.

Built a filterable internship search database in React, TypeScript, and Node.js, collating 200+ listings from LinkedIn and Indeed, containerized with Docker over an SQLite backend. Implemented Firebase OAuth authentication with Captcha and email verification security checks.

Designed and implemented a scalable ETL pipeline in Python to extract, clean, and transform heterogeneous clinical records across 46k+ diabetes patients from MIMIC-IV, engineering features for downstream unsupervised segmentation models (K-Means, UMAP) to identify clinically meaningful clusters.

A custom map that displays park information and allows Monterey County residents to search parks in a radius, by name, by hours, alphabetically, by park features, and give public transit routes.

Implemented a feedforward neural network (CNN) with 128-unit and 256-unit ReLU layers that classified inputs across 29 categories on a live camera feed input; Used binary cross-entropy loss & SGD to achieve an AUC of 0.97.