Built large language models (LLM) that moderate content and assess how content may violate a platform’s (ex. social media, chat sessions, etc.) guidelines (hate speech, doxing, etc.) with Python, LangChain, OpenAI API, Pandas, Pinecone, Scikit-learn.

Researched generative AI methods in content moderation and designed pipelines with Dr. Theodoros Evgeniou. Implementing techniques such as retrieval augmented generation, k-nearest neighbors, prompt engineering, and synthetic data generation.

Designed and developed Sietch, a data load-balancing backend service built in Java that interacts with various AWS network services and implements event streaming to send over 50 million metrics per second to a Kafka queue to fan-out to services; potentially saves $28k per month, a 47% reduction in cost.

Wrote Smithy model that defines service operations, data structures, variables, and HTTP bindings to implement type-safety, helping Sietch to be more secure and efficient. Self-taught multiple internal technologies such as Coral and Smithy.

Developed clustering algorithms for equity derivatives in Goldman’s proprietary language (Slang), leveraging risk data, hierarchical clustering, K-Means, and similarity metrics to produce potential clusters of risk factors that help to streamline the investigation of risk models that underestimate risk and captures risk factor correlational deviation.

Implemented backtesting on credit risk models to assess their performance and examine potential issues.

Presented work to 5 Managing Directors as well as other senior executives.

Worked with Dr. Diyi Yang (Social and Language Technologies Lab).

Built machine learning models to predict if a conversation on Reddit is civil or uncivil (Python, BERT).

Analyzed 20 million Reddit comments to investigate how different linguistic features (aspersion, provocation, vulgarity, stereotypes, etc.) correlate with civility and affect conversations and how incivility shapes the subsequent participation of the user in a community and opinions about a specific topic.

Presented to Meta research scientists. Published research paper “A Crisis of Civility? Modeling Incivility and Its Effects in Political Discourse Online” in ICWSM 2024.