iOS · production ML · cloud · systems
Software Engineer
I build and ship software end to end — an iOS app on the App Store, production ML pipelines on Azure, the analytics platform a company runs on, agent tooling and a mobile game. I pick the stack the problem needs and take it all the way to people using it.
Production document-intelligence pipeline - classification and field extraction over inbound clinical documents on Azure.
An iOS cooking app on the App Store — scan a grocery receipt, get a pantry, and ask an assistant what to cook with it.
Shipped to the App Store
An agentic second brain - a Claude-operated vault, custom skills, and a dashboard that runs my week.
Semantic segmentation of seafloor imagery + bathymetry across 17 sediment classes, U-Net with ResNet backbones.
IoU ~0.74 on held-out set · 17 substrate classes
Rather than describe the computer-vision work, here is some of it running. Draw something — the classifier is a small CNN I trained on Quick, Draw!, exported to ONNX, executing in your browser.
cat · fish · house · tree · car · star · sun · bicycle · umbrella · eye
A small convolutional net, trained on Google's Quick, Draw! data, exported to ONNX and run entirely on your machine — no server sees your drawing.
snapshot · 2026-08-17
This is a personal portfolio website built with Next.js and deployed via GitHub Pages. It showcases machine learning and software engineering projects, skills, and experience. The site is designed to be fast, responsive, and developer-friendly, using modern frameworks and build tooling.
Astro
gymsocial is a social fitness app for iOS that lets users log workouts, track muscle group activity, and share progress with friends. Built with SwiftUI and Firebase, it combines workout logging with social features to create a motivating experience for lifters.
Swift
This project presents a BERT-based deep learning model for sentiment classification on social media data from X (formerly Twitter). It fine-tunes a pretrained BERT encoder to accurately classify tweets into one of three sentiment classes: positive, negative, or neutral.
Jupyter Notebook
This project implements a deep learning pipeline for semantic segmentation of seafloor images using a U-Net architecture with a ResNet backbone. The model predicts 17 classes of sediment and substrate types based on RGB images and accompanying bathymetry data.
Jupyter Notebook
This project implements core multiresolution image processing techniques using Gaussian and Laplacian pyramids, including image reduction, expansion, reconstruction, and blending. It also performs seamless image mosaicking using pyramid-based blending.
Python
This project explores image classification using convolutional neural networks (CNNs), including both custom-built architectures and transfer learning with pretrained models. The experiments are conducted on the CIFAR-100 dataset and demonstrate the effectiveness of CNN-based feature extraction for image recognition.
Jupyter Notebook