Currently building software to assist with regression testing of camera algorithms such as 3A using cloud tools, emulators, and raw or simulated sensor data. I also help research novel camera algorithms that help make images that are good for self driving purposes, not specifically human perception purposes.

Projects:1. Worked on team of 3 to write firmware for custom cameras for cruise. The camera utilized the CV22 ISP provided by Ambarella. The programming environment consisted of a ThreadX RTOS instance for interacting with the ISP and a Linux instance to communicate with the RTOS and implement streaming. Primarily involved in providing tools for manufacturing, light RTOS work, controlling the camera settings, and writing the streaming portion of the firmware for the cameras.2. Created framework for validating sensor firmware versions across the car, along with tracking serial and part numbers across the fleet. The car has 20+ sensor instances (so multiple of the same sensor in some cases) and new firmware was constantly being flashed onto individual units. My framework helped verify that all of the sensors on the car had the correct firmware versions for each car. This increased fleet reliability by preventing cars with the wrong firmware versions from being deployed on drives. During this project I also helped build other "prelaunch" checks that made sure the network was in a good state using SNMP and also come up with a schema for uploading car launch errors to a database to allow the errors to be easily queried in big-query or looker.3. I also worked on lidar integration into the stack. Cruise had purchased the Hesai P64 lidar and I wrote the ROS driver for integrating the device into the stack. I also wrote monitoring code to make sure that the device was correctly configured on start up and monitored the state of the device while running (to catch issues like overheating or RPM threshold violations). I did similar work for integrating an Ouster lidar into the car stack.I've also done a lot of general C++ work to help people modernize their code.

Worked on tools and libraries used to support core features of the PI3 server. Last Project:Led group of 5 to do a major refactor of critical subsystems in the PI System for performance and reliability improvements.Previous projects:Small library for remote administration of Windows machines written against .net standard. It is extensible and can be used to dispatch commands over WinRM or SSH.Worked on guarding against critical data loss/corruption of memory mapped data queue files in case of power failure or hard shutdown of a machine.Worked on tools to provide customer more feedback/insight into connection history of the PI Server for debugging purposes. Reduced build time of code from potentially hours to minutes using NuGet and PowerShell to precompile and distribute widely used internal libraries and custom builds of the C++ Boost libraries. Built a library to create signing certificates and certificate chains to simulate a certificate authority. This was for testing our X.509 support and seeing if we could handle different use cases. I modeled the chains as a directional graph using Boost.Graph and took the graph metadata and used it to construct the certificate chains through the CryptoAPI library. The certificates were linked through Authority Key Identifier and Subject Key Identifier chaining. The graph could also be exported to graphviz for visual verification of correctness.Integrated TLS and X.509 support into the PI Server.Continuously working on refactoring and improving code base to be more safe and performant (examples include template metaprogramming, better memory management through smart pointers, and usage of std::algorithm and boost libraries to remove outdated and error prone internal versions of standard types and algorithms). I also spend a lot of my time unit testing.

Worked on Google Chrome and built a framework to test its Print Preview.Built a sophisticated test that can control Chrome to navigate to a web page, print it to PDF, then convert the PDF to an image file.Extended the WebKit Layout Testing Framework to run the test and compare the image the test generated with a reference image. If the images are very different, the comparison will fail, meaning that there has been a regression or an improvement in how Chrome's Print Preview generates PDFs.The test framework was deployed on 3 test machines that run the tests every time a change is made to Chrome. It actually tests real scenarios and has exposed certain flaws in how Chrome renders the Print Preview PDF, especially on Windows.

-Helped fix some critical bugs and created unit and functional tests for these bugs-Helped test several data structures that used Boost::ASIO and C++11-The testing helped lead to successful prototypes of these data structures-Created a GUI using Winform and C# that pulled data from Microsoft Team Foundation Server and allowed the user to select test suites and run them against specified builds overnight

I was assisting a PhD student in his research about human robot interactions. Specifically I was assisting him with technical problems, such as solving dead zone issues that are present when using multiple kinects at many angles and large distances to create point clouds of people.

I was a counselor in training at a camp and also helped teach about 150 to 200 children, ages 4-15, how to float and how to swim multiple swimming styles.