Selected work
A mix of applied AI, optimization, visualization, and production-oriented systems work.
The portfolio leans toward projects that show how I think through real problems: building models, shaping decision support, and turning experimentation into something practical.
Product backend tooling
This private product backend supports an image-to-canvas workflow that transforms uploaded photos into palette-constrained previews and print-oriented guides. The writeup focuses on architecture, API design, image-processing boundaries, storage abstraction, and dashboard review without exposing private fulfillment details.
Technologies: Django, Python, REST APIs, image processing, cloud storage, webhook orchestration.
Read MoreHealthcare machine learning
This project compares classical classifiers and neural-network experiments on the Breast Cancer Wisconsin Diagnostic dataset. The case study emphasizes feature filtering, model comparison, ROC-style review, and diagnostic decision-support framing without implying clinical deployment.
Key Skills: Data Analysis, Neural Networks, Medical Diagnostics, Feature Engineering.
Read MoreAnalytical storytelling
This project investigates the environmental impact of international trade in OECD countries. Using OECD data, it examines carbon emissions embodied in both imports and exports and frames the results through a policy-oriented narrative.
Technologies: Data visualization, environmental analysis, international trade data.
Read MoreOptimization and public policy
This decision-analytics project translates Washington State redistricting into a constrained optimization model using county population, demographic, and adjacency data. The page now highlights both the modeled map and the tradeoffs that remain under county-level assumptions.
Technologies: Decision analytics, algorithmic optimization, geographic data analysis.
Read MoreComputer vision
This computer-vision capstone explores an Automatic License Plate Recognition pipeline across 433 vehicle images. The case study shows the detection-to-OCR workflow, confidence review, exact-match audit, and honest failure modes around lighting, crop quality, and plate formats.
Technologies: Computer vision, OCR, YOLOv8, OpenCV, Python.
Read MoreDeep learning
This deep-learning project uses convolutional neural networks to classify animal images, with attention to image preprocessing, model iteration, and benchmark-style evaluation.
Technologies: TensorFlow, Keras, CNNs, image data pipelines, Python.
Read MoreNLP and education
Walter is a domain-focused chemistry chatbot prototype backed by a 118-element reference corpus. The improved case study frames it as a grounded NLP proof of concept, with retrieval workflow diagrams, example responses, corpus coverage, and clear limitations.
Technologies: NLP, semantic retrieval, Python, sentence transformers, cosine similarity.
Read MoreOptimization and real estate
This project presents an optimization framework for evaluating multifamily development opportunities by combining land-use analysis, predictive modeling, and feasibility-oriented decision support.
Technologies: Optimization, linear programming, predictive modeling, Python.
Read MoreVisual communication
This collection brings together interactive and notebook-based visualization work across temporal, categorical, and hierarchical data, with an emphasis on analytical clarity and narrative structure.
Technologies: Data visualization, exploratory analysis, notebook workflows, interactive web graphics.
Read MoreBenchmark NLP
This NLP project classifies short social-media posts related to real disasters, combining text preprocessing with recurrent and transformer-era modeling experiments in a benchmark setting.
Technologies: NLP, LSTMs, BERT experimentation, NLTK, spaCy, Python.
Read More