Brief
At NVIDIA, I work with an amazing team of engineers to develop PLDM spec-compatible bundles. These bundles include the firmware used by every GPU chip on a rack to handle AI compute operations. My role involves managing the lifecycle of firmware packages, which essentially act as the low level operating system for GPU racks used by companies worldwide to train and inference their AI models. Together with my team, I help create the mechanisms that combine firmware components into a single package, which is then flashed onto the GPU racks.
Another big part of my job is creating and managing workflows to automate delivery of firmwares to Cloud Service Providers. I’ve built automated pipelines to streamline this process. As part of the firmware development team, I focus on overseeing the entire process of building and distributing the software that powers GPUs. The main goal is always to deliver the highest quality software for the world’s most advanced computing systems.
Technologies Used
- Python 3
- Jenkins
- PLDM Spec
- OpenBMC
Approach
To make sure the latest firmware is always ready with proper testing, we set up a centralized internal repository where teams can push their firmware. We also developed a way to fetch the latest versions whenever needed. This approach covers all the firmware that needs to be delivered to clients. We created an app to automate fetching and validating firmware. It runs sanity checks on a pool of local machines, tags the builds, and then releases the validated firmware to customers. This has made the whole process faster, more reliable, and easier to manage.
Reflection
Working at NVIDIA was my first time being part of a hardware-focused company, and I was given a big responsibility to work on client-facing projects. It wasn’t just about adding new features—I also had to follow strict gold-standard coding practices. I really enjoyed building new features and improving the codebase’s quality for the whole team. Overall, it was a rewarding experience that helped me grow both technically and professionally, especially in understanding what it takes to deliver robust, client-ready solutions.