At PolarStar, most of my work focused on improving internal operational tools and building dashboards for day-to-day use across the business.

I design and develop convolutional neural network models for detecting, segmenting, and classifying lung nodules in CT imaging. The goal of this work is to support earlier diagnosis and better treatment planning for lung cancer, while keeping the models accurate and reliable enough for clinical use. The work sits at the intersection of deep learning research and practical engineering. Models are developed with an eye toward real-world constraints, not just benchmark performance.

To make these models usable across the team, I set up and maintain our deep learning server environments. These environments are containerized so models can be trained, tested, and deployed in a consistent way across development and production systems. This reduces setup friction, avoids environment-specific bugs, and makes it easier for multiple engineers to collaborate on the same code and datasets.

I also lead the organization and construction of the medical imaging datasets used for training and evaluation. This includes curating diverse, high-quality CT data, managing labeling and preprocessing steps, and designing data pipelines that are easy to extend and audit. The focus is on building datasets that produce models which generalize well and behave predictably when exposed to real-world clinical cases.

Brainforest is an entity that I use to try out new things and perform ad hoc services. It provides me perspective in business operations, management, and accounting. It is by means no a finished product but keeps me busy and curious in my free time.

I have worked with Cayman Code Specialists to assist in report generation and automation using Google Cloud Platform for third party inspections. This has been valuable for me because it keeps me active with general engineering and how things work systematically. It’s important to understand what shelters you from the nefarious bugs and relentless utraviolet radiation …

In software development, I worked on designing and maintaining both automatic and manual route planning algorithms for in-airway and transthoracic navigation. These algorithms supported precise, minimally invasive procedures and had to be reliable under a wide range of clinical conditions. A large part of this work involved restructuring core library modules to better separate public, private, build, and install interfaces. This made the codebase easier to reason about, safer to extend, and less painful to maintain over time. I also handled exporting thoracic segmentations as DICOM-RT objects, implemented geometric post-processing to detect landmarks in CT images, and enabled background case processing so users could continue working without blocking on long-running tasks.

To modernize the development workflow, I rewrote the build system to move from a directory-based CMake setup to a fully target-based approach. This reduced coupling between components and made dependencies clearer and easier to manage. As part of the same effort, I upgraded the development environment from Visual Studio 2015 to 2022. This improved build reliability, tooling support, and long-term maintainability of the project.

On the systems engineering side, I worked on integrating key hardware components into the platform. This included Matrox Orion HD and Clarity UHD capture cards, as well as NDI Vega and Polaris tracking systems. The goal was to ensure reliable performance and clean integration with the software stack. I also developed internal engineering tools to test, verify, and integrate new features into our core algorithms. In addition, I built utility programs used by field teams to extract and analyze complex patient cases, troubleshoot issues, and manage equipment calibrations and registrations outside of the lab environment.

In development operations, I helped scale the team from a small local group of four to a global team of roughly forty. This involved creating standard operating procedures and development protocols that could work across time zones without slowing people down. I led the migration of our self-hosted GitLab infrastructure, including repositories, issue boards, and documentation, to GitLab.com to improve collaboration and reduce maintenance overhead. I also maintained software builds for product releases and acted as the administrator and maintainer for our GitLab groups and projects, making sure version control and release processes stayed predictable and usable for the entire team.

For EE 350 – Continuous-Time Linear Systems, I hosted two of six weekly recitation sections supporting lectures for 130 students and contributed to exam creation, supervision, and grading.

For EE 455 – Digital Image Processing, I led MATLAB tutorials, held office hours to support students, and handled grading responsibilities for a class of 60 students.