In the latest Developer Impact series, Ampere® Computing’s Dave Neary speaks with Dr. RJ Nowling from the Milwaukee School of Engineering to discuss how the school is bridging the gap between theoretical machine learning (ML) and real-world production systems and how students learn to build “real” ML systems that work on everyday software, not just smart mathematical models.
Dr. Nowling explains that many schools teach students how to design intelligent computer models. MSOE tries to go one step further. The key idea is: a model is not useful if it only works in a laboratory. In the real world, a working system also needs other parts, such as fetching new data, connecting to databases, converting raw data into the kind of numbers the model can use, and keeping track of whether the model continues to perform well over time.
He compares it to his previous work in the online advertising industry. There, the model itself was only a small part of the whole process. The system had to constantly incorporate new data, update information periodically, and check the “latest” information because that is often best for making quick decisions. Additionally, models can “drift” and become less accurate as the world changes, so the system must periodically retrain and update the model. That broader set of tasks is what makes machine learning production work difficult and important.
In this course, students practice building the entire system from start to finish. They configure databases (such as PostgreSQL), create data pipelines, train models offline, and then release them as services so other software can use them. They also learn how to monitor the model, notice when performance changes, and use automated tools to have the model retrained and redeployed when necessary. A big goal of teaching is to allow students to see real operational problems, such as running out of disk space or having problems with containers, so that they learn to troubleshoot like engineers do in the real world.
The video also explains why Ampere computers are part of the training. Dr. Nowling says they run many student lab environments on a single Ampere server, using CPU power instead of relying on GPUs. Students typically share the machine via virtual computers, so each can run their own configuration (databases, pipelines, services, and automation jobs) at the same time.
This is important because many people assume that “AI always needs a GPU.” Dr. Nowling says that’s not always true. For many practical tasks, such as quickly detecting fraud or deciding whether to show an ad, speed is important and CPU-based approaches can work well. The school wants students to understand those real-life limitations, not just the most famous AI setups.
It also talks about how the software works smoothly on this system. They use a common Linux operating system (Debian) and tools to configure everything, so that students are not stuck struggling with software problems.
Finally, the program is also expanding into generative AI (such as tools that can produce text). Students learn that real-world AI systems often combine different “smarter but smaller” models rather than using one huge model for everything, which can reduce costs and improve reliability.
Watch the full video here:
To learn more about the efforts of Ampere developers and find best practices, visit Ampere Developer Center and join the conversation on the Ampere Developer Community.
Source link
