Update cv

Author: aurora-admin

content/cv.md

@@ -35,7 +35,7 @@ title: "Curriculum Vitae"
     <li><p>I led the design and development of bespoke real-time vision inspection software for various manufacturing processes. Some examples include:</p></li>
     <ul style="margin:0;">
         <li><p><a href="https://www.ussvision.com/post/in-line-split-detection">A defect detection system</a> for inspecting complex sheet metal panels (primarily car parts) originating from a large stamping press. Originally designed for detecting splits as small as one millimeter in the metal, the system has been expanded for verifying hole presence, as well as wrinkle and galling detection.
-        <br/> With thirty GigE cameras, the system processes sixty to ninety 12.3MP images every second, running the data through a custom-made processing pipeline which consists of both traditional signal processing algorithms as well as an ensemble vision transformers and variational autoencoders that were developed in-house. <br/> This was a distributed system which required a significant amount of computation; the main processing engine included a multi-socket motherboard with several AMD EPYC processors, three Nvidia A100 GPUs, over 1 TB of RAM, and a 40 GiB/s NIC. This has been installed on five stamping presses belonging to various automotive OEMs with an average of a 10x reduction in outflow across all.</p></li>
+        <br/> With thirty GigE cameras, the system processes sixty to ninety 12.3MP images every second, running the data through a custom-made processing pipeline which consists of both traditional signal processing algorithms as well as an ensemble of vision transformers and variational autoencoders that were developed in-house. <br/> This was a distributed system which required a significant amount of computation; the main processing engine included a multi-socket motherboard with several AMD EPYC processors, three Nvidia A100 GPUs, over 1 TB of RAM, and a 40 GiB/s NIC. This has been installed on five stamping presses belonging to various automotive OEMs with an average of a 10x reduction in outflow across all.</p></li>
         <li><p>A thermal imaging application to detect anomalies in lithium-ion battery temperature, alerting the factory to a potential fire or explosion hazard before it occurs. This uses several thermal cameras to rapidly inspect all parts of the battery as they move into an inspection station, with surface pixel error margins below 1°C.</p></li>
         <li><p>A quality checking system to inspect laser engravings on vehicle headlamps presented to the camera by a robotic arm. This required significant image preprocessing due to small variations within the presentation angle causing significant variation in the appearance of the engraved characters. As a result, several stages of non-linear contrast enhancement, thresholding, and morphological operations were applied based on parameters estimated from the measured intensity of the laser engravings before utilizing an OCR engine to verify the characters were correct.  </p></li>
         <li><p>For each system, we provided web dashboards on servers that we personally maintained in order for both our team and the client to view metrics and captured images stored in databases, as well as annotation tools for automated model training and feedback.</p></li>