Software engineering / Applied AI / Systems
Colby CS student, expected May 2028. Preparing for Summer 2027 SWE internships, with experience across LLM agents, analytics platforms, AI health tools, and market simulations.
Looking for Summer 2027 SWE internships on teams building backend infrastructure, applied AI products, developer tools, fintech systems, or full-stack applications with real users.
Colby College, B.A. Computer Science, Math minor. GPA 4.12. Coursework includes Programming Languages, Algorithms, Computation Theory, Game Theory, and Probability.
Shipped an agentic extension for audience targeting in Premium A/B testing. Built an LLM-to-DSL compiler, an evaluation framework for syntax and semantic validation, and a LangGraph orchestration layer routing requests across 10+ agents and API services.
Deployed a real-time shopping analytics dashboard across 50 global retail stores, enabling $1.6M projected annual revenue uplift. Refactored GraphQL mutations to remove duplicate Apollo calls and built a cross-store search tool that reduced lookup time by 26%.
Trained neural networks for loan outcome prediction, then built decision-tree surrogate models from model outputs to explain decision boundaries. Improved human prediction accuracy by 32% using model predictions and interpretable rules.
Worked with a pre-seed health tech team to improve an AI dementia diagnosis API. Containerized the model with FastAPI, deployed on AWS ECS, reduced inference to 2 seconds per diagnosis, and used SHAP explanations to cut 4 hours per week of manual verification.
Built an LLM-based Alzheimer's analyzer that converts audio into structured transcripts and diagnostic summaries. Optimized RAG retrieval, transcript preprocessing, and LLaMA prompt context to reduce inference cost and clinician review time.
Central limit order book with O(1) best-bid/ask lookup, O(1) cancellation via hashmap, Hawkes process order arrivals, Kyle's lambda, VPIN, OFI, and event-driven backtesting.
Full-stack market intelligence dashboard with stock data, news sync, and AI-assisted market context.
PyTorch CNN and Streamlit app for classifying German traffic signs across 43 classes.
Natural-language and SMILES-based molecular search with embeddings and interactive graph exploration.
Nguyen, X.K., Ngo, T.A., et al. "A Comparative Study of Transfer Learning Techniques with BERT Family Models in Early Diagnosis of Alzheimer's Disease." IFMBE Proceedings, vol. 122, Springer Nature, 2025.