Matrix Design Group

Computer Vision Engineer Intern · May – August 2024

Spent the summer building computer vision tools and infrastructure for a small engineering firm that does interesting work with CV object detection.

What I worked on

Synthetic data generation for CV training
Experimented with generating synthetic training data in Unity (C#) to improve object detection models. Managed to bump the effective detection rate by 2% in an experimental setting.

Data pipeline infrastructure
Built Python tools for training (via PyTorch), metadata documentation, and dataset reorganization. Integrated Voxel51 for image quality assurance and created a pipeline to transform video files into high-quality, deduplicated datasets. This streamlines dataset curation for the team.

Desktop application development
Worked on the frontend and backend of a .NET/C# desktop app for data import, filtering, and visualization. Connected it to MySQL for storage and retrieval.

What I learned

This was my first real exposure to computer vision in a production environment. Got to see how CV systems actually get deployed, not just trained in notebooks. The team was small enough that I could work on multiple projects.