Astrobotic Technology | Embedded & SW Engineering | Pittsburgh, PA | Onsite | Full Time | US persons only | https://www.astrobotic.com
Astrobotic is a small space robotics company with big ambitions. Our R&D team has numerous contracts to develop and deliver in-space navigation and in-space computing software and hardware to help spacecraft land on the Moon and other planetary bodies We need multiple talented robotics/software engineers to support and take leading positions in the development of our technology portfolio. We’ll be developing novel sensors and flying them in-space within a few years.
Our R&D team operates autonomously and collaboratively. We hope that new researchers will be eager learners with ambitions to master their field in a short period of time by challenging themselves and aiming high. While there is a lot of work to do to deliver on our technical contracts, we hope to find people who believe they can dream and develop new technologies to increase our access to space.
I messed around with some slightly above toy work, implementing expectation-maximization for gaussian mixture models and some other supporting tools for 3D point clouds, and found it really intuitive for someone with a decent amount of FP experience but little GPU experience.
Biggest challenge for me was a lack of profiling tools when running OpenCL on an NVIDIA GPU, I got OK performance but had no way to identify hotspots or opportunities for improvement.
If anyone has AIAA paper access (or a workaround ;), they published a ton of details on their guidance, control, and avionics designs at the beginning of this year:
Very interesting, I recently put together a simple caffe static compiler of my own on the path to FPGA deployment of CNN inference pipelines. I looked at halide[0] as a possible intermediate representation for CPU, GPU, and FPGA, (as a step above LLVM IR), but Likely seems like an interesting option.
I don't think you can do high frequency RAM lines with hand etched boards or BGA escape routing without vias, but two layer prototype boards are pretty cheap these days.