• Architected an LLM-based chatbot for searching through internal company documents and answering queries with generated factual responses using Python, Event-Driven architecture, and AWS components, slashed engineers’ information retrieval times and drove department-wide adoption.

• Created a Graphql Spring server in Kotlin to trigger Temporal Workflows. Deployed via Harness onto AWS ECS instances and monitored traffic via distributed tracing using Datadog APM.

• Coordinated with compliance and investment stakeholders to build a data ingestion pipeline and scheduled jobs for securely fetching customer SSNs to satisfy FINRA CAIS regulatory requirements, all within a short 2 month notice.

• Built a data pipeline of 28 different models using Airflow, DBT, and Datadog to monitor and alert on invalid customer states, exposing incidents requiring time-sensitive remediation.

• Led a research project on the Ego4d team in Meta AI Research to perform object detection on a

video dataset of robot arm movements for typical kitchen actions with different camera perspectives

• Master's thesis supervised by Muriel Médard in the MIT Research Laboratory of Electronics department

• IEEE International Conference on Acoustics, Speech and Signal Processing 2023 paper submission accepted: "InfoShape: Task-Based Neural Data Shaping via

Mutual Information" (https://arxiv.org/abs/2210.15034)

• Designed and implemented a novel Lagrangian optimization scheme for a neural network encoder with a mutual information estimator, ReMINE (Choi et. al.), using Pytorch, NumPy, and sklearn

• Empirically demonstrated InfoShape's ability to shape synthetic data to achieve both a 35% decrease in privacy leakage and a utility score increase by a factor of 5.1