i go by john, a senior at the University of Notre Dame, double majoring in cs & applied math. broadly, i work on trustworthy NLP, focused on interpretability of language models. i research closely with nuno moniz and meng jiang.
previously, i was an ai/ml sde intern at AWS SageMaker Unifed Studio where i built, deployed, and scaled AI agents. i worked end-to-end, including infrastructure (IaC), to agentic workflows with memory/hitl, and MCP/A2A integrations.
Selected works
Projects
Open-source contributor for RL optimizations, Arbor
Built NVIDIA-GPU monitoring tools for training, JQlang query logging for debug, and used vLLM for reinforcement learning. Added CLI/Python module for config management, API startup, and example scripts.
Built bidirectional LSTM encoder-decoder architecture with multi-head attention mechanism for seq2seq learning. Pre-processed equations to LaTeX, then corrected with LLM using in-context learning with 92% validation accuracy.
1st Place Advanced Databases Final Project ($1050)
Architected React Native app with Expo/Redux, Python (FastAPI), Oracle database, AWS EC2 hosting and S3 buckets. Coded collaborative filtering recommendation system using PL/SQL procedures, indexes, views and job scheduler.
Built entire backend with Flask, PostgreSQL (pgvector for embeddings), Redis/Celery for task scheduling, AWS hosted. Generated 10k synthetic trips, automated clustering/vectorizer model updates, designed content-based recommender.