Project1 The AI Math Assistance Calculator, attempted to provide students, or the user, with viable help in solving ordinary or complex mathematical operations. Specifically, for this project, we focused on training the GPT-3 model with the AP Physics test in mind. We choose to train the model with this in mind, to prove a use case. Eventually, this model can be tuned to aid accurately in solving many different subjects. The user can input any question from an AP Physics examination or practice problems, and the model will provide immediate steps on how to go about solving such problems. We focus on tuning this model to not give numerical solutions as a whole, as we wanted the model to assist in learning, not necessarily solving. Although this has its most efficient use as a web application, where students could input entries. We opted to have it in the form of a calculator, allowing the user to interact with the software in an interesting and meaningful way. This project demonstrates proficiencies in both software and hardware. All the hardware, besides the actual microcontrollers, was custom designed for this calculator, as well as the firmware necessary to run all peripherals. A completely custom enclosure was also printed, using SLA technology.
Project2 The lidar scanner, as a method to utilize computer vision to re-create 3D models and classification, showcases the use of the continually changing technology of computer vision and its capabilities. This project delves deep into both the software and hardware necessary to provide meaningful information through computer vision. The project's goal was to scan and remodel a physical object while classifying it as well. The classification was kept simple, in this case, it could determine between a cube and a cylinder. This allowed for common testing of regular objects, such as tissue boxes for cubical shapes, and mugs/containers for cylindrical shapes. To collect data, a lidar system was utilized, which allowed for a point cloud collection of data. To perform the scan, the hardware had to be envisioned which allowed the correct collection of data points. A simple rotational platform allowed for the data points to be collected on the same y plane, but different x points. The lidar system itself, utilized translational movement on the y axis, to change the y plane, allowing for a complete collection of data points. The project was successfully able to model physical objects from collected data, while also performing classification, being able to determine the difference in shape between a cylinder and a cube. This project demonstrates the combination of both hardware and software skills necessary to successfully create a computer vision system from scratch.
Project3 The microwave leakage project aimed to demonstrate a basic form of RF energy collection or harvesting. By capturing transmitted RF signals, the device could directly power a low-power circuit, in this case, an LED. Researching voltage multipliers, I stumbled upon the Greinacher circuit, which provided reduced ripple compared to simpler circuit designs, like the Villard Circuit. Reimagining this circuit design, it was realized for this specific use case. This design was integrated into a custom PCB, which allowed for easy integration with an antenna. For the goal of detecting microwave leakage, an antenna suitable for a frequency range of 2.4-2.5GHz was determined. This project demonstrates the research and implementation to create hardware necessary for a specific reason, in this case detecting microwave leakage.