Adrian Apsay Email & Phone Number

United States

• Developing an information retrieval system in Retrieval-Augmented Generation (RAG) architecture for email data sent to incarcerated individuals, leveraging OpenAI for embeddings, and Azure AI Search and Cosmos DB for vector search and indexing to extract relevant information, detect anomalies, and generate insights based on client queries• Implementing clustering algorithms (K-Means, HDBSCAN) on 2024 email data, utilizing vector embedding representations and employing UMAP, Plotly, TF-IDF, and the OpenAI API to visualize and topic model each cluster as well as Python, scikit-learn, and Jupyter Notebook for further exploratory data analysis• Engineering Extract-Transform-Load (ETL) data pipelines to migrate and transform data from on-premises SQL Server databases to data lake cloud storage using Azure Synapse Analytics, PySpark, and T-SQL, contributing to the department’s early transition to scalable, cloud-based data workflows

La Jolla Shores, California, United States

[(3x - WI24, SP24, FA24): DSC 20 - Programming and Data Structures]• Applying a strong understanding of Python to competently answer 500+ students’ inquiries during office hours and in the online class forum regarding programming, data structures, time-complexity, and object-oriented programming• Hosting office hour sessions to provide instructional guidance to students, receiving 100% satisfaction rate• Diligently grading programming assignments and exams to ensure proper and accurate grading

La Jolla Shores, California, United States

• Developed and implemented deep learning models for speech emotion recognition using the CREMA-D dataset, achieving a 20% improvement in accuracy with our CNN model and a 32% improvement with a HuggingFace pre-trained transformer, significantly outperforming the accuracy of human interpreters• Collaborated with three undergraduate engineers to build a Convolutional Neural Network (CNN) using PyTorch, utilizing Mel-Spectrograms from .wav files for feature extraction in emotion classification• Optimized hyperparameters using Optuna and continuously evaluated performance metrics via scikit-learn, addressing the challenges of subjective emotion interpretationFeel free to take a look at our Project Team's app and project poster below.

La Jolla Shores, California, United States

• Formulated an in-depth curriculum for the Summer 2023 Boot Camp using Google Colab and Slides; hosted by Diversity in Data Science and cosponsored by HDSI and TILOS• Instructed high school students at a Data Science summer boot camp at UC San Diego, involving components of Python programming, data analysis, visualization, and the Data Science Life Cycle using Google Colab• Educated vital Python packages (pandas, NumPy, Matplotlib) and utilized them to teach core Data Science principles through hands-on experience with datasets

Bachelor Of Science - Bs, Data Science

Bachelor Of Science - Bs, Data Science

Activities and Societies: Provost Honors, Association of Computing Machinery (ACM), Data Science Student Society (DS3), Diversity in Data.

High School Diploma

Activities and Societies: Associated Student Body, National Honors Society, Link Crew, Sociology Club, High School Volleyball4.59/4.00.