Was on a 5-person tiger team that improved the overall performance of our AMRs by 135% over a 6-month period. As part of this effort I designed the obstacle persistence functionality to ensure safety when traveling at higher speeds and improved the efficiency of our lidar filtering pipeline to reduce the CPU by 85%.

  As tech lead I identified several focus areas that were impacting our ability to produce and deployment a robust product. These focus areas included a lack of calibration procedures for our robots, a testing strategy that was letting too many bugs make it to a client site and a lack of simulation tools which made it difficult for dev engineers to thoroughly test software changes. To address these short comings the following actions were taken:
     * Wrote automated calibration scripts and worked with the manufacturing team to provide training and incorporated feedback on how to improved these scripts.
     * Worked with the validation team to define a new nightly test mock run that more closely resembled workflows from real customer sites and increase the test from a 20-pallet/2 robot shuffle to a 43-pallet/6 robot shuffle. This has significantly increased the miles driven by our robots on each release candidate increasing our confidence in the relability of the software before shipping to customer sites.
     * Produced a proof of concept demo of a simulation tool that matches our current products. This tool has been adopted by members of the team to increase testing efficiency and has shown enough value to management in order to dedicate multiple people to work on improving this tool further.

  Spent time working closely with the deployment team both on site and remotely analyzing and improving the performance of our robot fleet at customer sites which produced several key results:
     * Analyzed data to identify system bottlenecks and improve pallet throughput by 39% at an automotive parts manufacturer
     * Addressed safety concerns and completed customer acceptance testing at a retail warehouse customer
     * Filtered LiDAR data to remove hits from dust to reduce robot stop time by 52% and support touches by 82%