Software Engineer II - Offboard Python Application

Latitude AI ( lat.ai ) is building the future of Ford’s autonomy roadmap to make travel safer, less stressful, and more enjoyable for everyone. Bringing this vision to scale, our fully in-house developed hands-free ADAS platform will debut on the all-new Ford Fathom in 2027. When you join the Latitude team, you’ll work alongside leading experts across machine learning and robotics, cloud platforms, mapping, sensors and compute systems, test operations, systems and safety engineering – all dedicated to redefining the relationship between people and their vehicles for millions of customers. As a Ford Motor Company subsidiary, we operate independently to develop automated driving technology at the speed of a technology startup. Latitude is headquartered in Pittsburgh with engineering centers in Dearborn, Mich., and Palo Alto, Calif. Meet the team: The Virtual Test Ecosystem (VTE) team ensures that our simulation (fully synthetic) and resimulation (replay engine using vehicle logs) workflows run efficiently and reliably at large scale. We use a mix of open source and cloud provider services to implement a processing framework that’s cloud-agnostic as well as capable of running on-prem. The results of these simulations is business-critical for us - they’re used in order to allow autonomy developers and testers to do large-scale safety and performance testing of changes prior to in-vehicle testing. We interact with engineers throughout the onboard and offboard stack in order to ensure smooth end-to-end capabilities. What you’ll do: • We run vehicle simulations at unprecedented scales in order to ensure we’re releasing a safe product. You’ll support this via improvements to our sim/resim cloud framework and applications that run on this framework. In terms of implementation, this involves a lot of Python coding, interacting with open source packages and various AWS/GCP services. In terms of analysis, we’re constantly looking for bottlenecks that limit scalability or otherwise hinder performance. You’ll put together benchmarking to identify these spots and associated dashboards to track and summarize the information. Then armed with that information, you’ll implement solutions to increase efficiency. • It’s not enough to have a framework that is functional - it needs to be intuitive and easy to use. We take our best guesses at this, but things evolve over time based on user interaction - both direct feedback as well as realizing how users are accidentally misusing or misinterpreting the results. VTE supports autonomy, systems, test, and a variety of other teams across the organization. Having an eye for how the system is being used, what users need, and a willingness to engage with these users on future improvements is critical. • We also can’t build it all, so we want to make the framework well-documented and as self-serve as possible for others to extend it (while being there to answer questions). For example, when there’s a new portion of the autonomy stack, we want that autonomy team to be able to update resim to add it in themselves so that they can iterate quickly. • In order to run logs through resim, we need to first know what logs are interesting. You’ll contribute to the ecosystem of how we manage storing and searching for these logs efficiently, including how we automate indexing the interesting actors and interactions in a scene and how we determine our dataset coverage. • Once we’ve run resim, we want to do something with the updated logs. Frequently this involves metrics to analyze the AV’s performance; these metrics could range from low-level performance of a particular detector in the stack to if the AV is meeting high-level business objectives. You’ll both work on the sim/resim cloud framework implementation to