Abby Seseri

WNBA Analytics Web Application

Built a full-stack WNBA analytics platform featuring a custom Scikit-learn model to find statistically similar players. The entire application is containerized with Docker and includes a CI/CD pipeline with GitHub Actions for automated testing.

Predicting Text from Intracranial Neural Signals

Optimized a GRU-based Brain-Computer Interface (BCI) decoder to predict text from neural signals using PyTorch. Implemented advanced regularization techniques, AdamW optimization, and time-masking augmentation to reduce overfitting. Achieved a Phoneme Error Rate (PER) of 18.37%, representing a 16.4% improvement over the baseline model.

AgentSociety: User Behavior Simulation Agent

Developed an LLM-based agent to simulate realistic user behavior using Chain-of-Thought reasoning and Iterative RAG. Implemented a modular architecture with smart context selection and user persona extraction to strictly enforce writing styles. Achieved a 5.06% improvement in review generation quality compared to the competition baseline.

Reliable, Secure Network Communication System

Developed a reliable, encrypted bidirectional data transfer protocol over UDP. Implemented AES-256 encryption, HMAC for authentication, and ECDSA/Diffie-Hellman for secure key exchange. Ensured data integrity and confidentiality using TLS-like handshakes and cryptographic standards.

HTTP Server and CRUD API

Engineered a multi-threaded HTTP 1.1 server in C++ supporting static resource serving and a JSON-based REST API. Designed a modular RequestHandler architecture using Boost.Asio and implemented a production-grade build pipeline with CMake, Docker, and GoogleTest for automated coverage and integration testing.