I'm currently a DevOps engineer at a Fortune 500 tech company, where I've worked for six years. My role focuses primarily on DevSecOps, and I manage a solid and easy-to-maintain CICD pipeline. Recently, I've been asked to venture into the ML-Ops space as part of my personal goals for the upcoming fiscal year. I'm expected to find relevant certifications on my own, but I feel uncertain about where to begin. I participated in an AI hackathon recently, helping teams with cloud setups and troubleshooting, but I'm primarily familiar with the basics of AI and not much more. Since I work for a major graphics company, I know understanding how to deploy different models, particularly involving Generative AI, is crucial. What's a good starting point for someone like me?
1 Answer
The ML-Ops field is really heating up. When we started our project, we had to navigate everything from model deployment to managing production costs. The certification options out there are all over the place, though! If you're looking to deploy models, I suggest you get a handle on the core concepts first—think about containerizing models, working with GPU allocation, and setting up model registries. Once you're comfortable with that, you can dive into A/B testing frameworks and monitoring for model drift. For closed model APIs, it's easier since you're mainly managing rate limits and token usage, but self-hosting requires attention to inference optimization and batching. If you're interested in AWS, their ML specialty certifications cover much of the deployment aspects but remember they're pretty AWS-centric.
Is it a chaotic challenge in terms of required skills, or does it feel like there's just a high demand for professionals in this area?
Thanks for the insights! I was thinking about whether it would be smarter to start by deploying to AWS using Bedrock instead of diving into local deployments right away. That way, it seems like I could focus on learning the essentials without getting bogged down by the nitty-gritty of optimization and batching at first.