Building algorithms quantifying how stylistically cohesive interior design products are and to select optimally cohesive populations of products subject to specific constraints.

Using concepts from the recommender systems literature to improve our product search implementations

Using self-supervised and fully supervised representation learning techniques to implement visual search

Working on everything from ML research and training to ML Ops to integrating my work into our frontend services - my stack includes PyTorch, Sagemaker, S3, Lambda, ECS, MySQL, Elasticsearch, backend work in Django, and frontend work in React

Conducting literature reviews and adapting convolutional neural networks for specialized segmentation tasks

Building a synthetic data generation pipeline to create novel rendered views of existing 3D room models and accurate ground truth data.

Designing and building a visual SLAM implementation to utilize captured pose data for better accuracy than available open source solutions.

Other work includes traditional computer vision tasks, optimization problems, backend pipeline code in Python, frontend work in JS + React, and experimental features fusing classification models with 3D spatial reasoning algorithms

Co-author on Modsy's pending patent for our room-building pipeline

Developed dynamic ingestion counter for display in company lobby.

Built internal Tableau proof-of-concept for democratizing data pulls.

Worked extensively with Amazon Redshift, Postgresql, Python, Javascript.