The Microstat (3D Clinostat)
I designed and built a tabletop 3D Clinostat for microgravity experimentation (deliverable as part of the greater NeuroCube project) using commercial-off-the-shelf mechanical and electronic components. The entire project is documented and open-source on GitHub.
The large axis of the Microstat rotates at 10 rpm , and the small axis at 4.5 rpm. It takes up roughly 350 x 300 x 320 mm of space when fully assembled and is run via a 12V, 5A power supply. It has been designed to operate from within most laboratory sized incubators up to 110 degrees Fahrenheit. For order inquiries and quotes, please contact me at david.krupp@uga.edu.
NeuroCube BioChamber
NeuroCube is a NASA-funded small satellite (6U CubeSat) experiment designed to study neuronal behavior in microgravity. I served as the primary Mission Specialist and lead architect for the experimental platform, coordinating between the Small Satellite Research Lab, the Center for Geospatial Research, and the Regenerative Bioscience Center at the University of Georgia. I designed the overall 6U satellite architecture along with the BioChamber, a compact pressurized enclosure for housing cell cultures in Low Earth Orbit, and developed a 2D gantry system for a moving Miniscope to capture high-resolution imaging of 25 individual neuronal cultures during flight. This project supports space biology research and contributes to understanding how microgravity impacts cellular systems.
Optimizing Neural Radiance Fields for Satellite Image Reconstruction through Solar-Informed Volume Rendering
As part of my research at the University of Georgia’s Small Satellite Research Lab, I, along with my team, developed an improved Neural Radiance Fields (NeRF) approach designed specifically for satellite image reconstruction. Traditional NeRFs rely on multi-angle, consistently lit images, which are rarely available in satellite remote sensing. To address this, our team introduced physics-based solar ray tracing and integrated the Blinn-Phong shading model into the rendering pipeline, allowing the model to account for the sun’s position and realistic surface reflectance. We tested this “Enhanced-Sat-NeRF” on real satellite data from WorldView-3 and compared it to Sat-NeRF and EO-NeRF baselines. Our results showed that Enhanced-Sat-NeRF produced comparable final reconstruction quality (mean PSNR ~27.2) while achieving more stable performance during early training stages, with reduced standard deviation in PSNR. Although final accuracy gains were not statistically significant, the early consistency suggests potential for better generalization across diverse terrains and seasonal lighting conditions. This work supports more accurate digital elevation models and urban mapping, with applications in land use analysis and biomass estimation in vulnerable ecosystems.
Vibration Table Controller
This section will describe my design and development work for a custom vibration table controller system. Details, specifications, and code repository links will be added here soon.