--- title: "Want to Support the SWEs You Work With? This Should Help" newsletter: "MLOps Community" date: 2025-03-18 source: https://aaif.live/newsletters/mlopscommunity/2025-03-18-want-to-support-the-swes-you-work-with-this-should-help --- # Want to Support the SWEs You Work With? This Should Help *Plus: AI adoption shifts, multi-agent swarms, SQL simplification, and rethinking LLM evals.* *MLOps Community — Agentic AI Foundation, 2025-03-18* Google has joined the likes of Celine Dion at Caesars Palace, David Guetta at Ushuaïa Ibiza, and of course, Elvis Presley at the International Hotel Las Vegas with their own residency [https://go.mlops.community/i427xq]. ## ML CAREERS We’re putting together a resource on what it really takes to progress in ML engineering - featuring insights from senior ML engineers who’ve tackled the challenges of scaling models, leading teams, and staying ahead in an industry that’s always evolving. Sign up here [https://gatewaze.mlops.community/offer/mlops-ml-engineers-careers-guide] to get updates as it comes together. And if you’re a senior ML engineer with lessons to share, we’d love to hear from you - drop us a message at binoy@mlops.community With that in mind, I'm really excited to share this collaboration withZack Schrag [https://www.linkedin.com/in/zackschrag/]! He’s been a software engineer for over a decade, currently at Meta, and regularly shares insights on career growth, problem-solving, and the human side of software engineering. This piece looks at the shift from software engineering to ML engineering - something many of you will have seen firsthand as AI adoption spreads. Whether you're deep in ML already or working alongside software engineers making the jump, it’s a useful perspective on how expectations are evolving and where your skills can bridge the gap. Zack’s also midway through a great series, “10 Things I've Learned in 10 Years as a Software Engineer” - worth checking out here [https://schrag.substack.com/archive]. FUTURE PROOFING YOUR CAREER WITH AI - A SOFTWARE ENGINEER’S PERSPECTIVE 📌 Note: the term AI has become increasingly associated with Generative AI and LLMs (Large Language Models). Here, it’s used to refer to AI as a whole–Generative AI and “classic” ML included. If you were a software engineer anytime since ~2015, there’s a good chance you were expected to be comfortable working with cloud platforms and cloud technologies (e.g. containers, Kubernetes, AWS, etc.). Do a search for “software engineer” jobs on LinkedIn, and the vast majority of them will include some cloud platform or technology on their list of requirements. Cloud isn’t the only technology that software engineers have been expected to adopt over time. When was the last time you saw a job posting for a Database Administrator? How about a Release Engineer? Database management and software release management are now baseline expectations of a modern software engineer–empowered by the improvement of tools, automation, and convergence of best practices. Roles like these follow a pattern: they come into existence to tackle the complexity of a specific technology, but over time the expectations tend to become absorbed into responsibilities for a software engineer. With AI, this pattern will likely continue. Rather than replacing software engineers, AI will become an expected part of the modern software engineer’s toolkit. According to Bloomberry, roles like AI Researcher, Applied Scientist, and Machine Learning Engineer have grown by 79.7% and 69.73%, respectively, from early 2023 to early 2024. This surge isn’t surprising given the AI boom sweeping the industry. But once cutting-edge models are built, how do they actually make it into production? That’s where software engineering comes in. Software engineers and DevOps professionals are increasingly working alongside data scientists and ML engineers to bring AI models from the lab to the real world. This collaboration means integrating traditional software best practices with machine learning workflows to ensure systems are reliable, scalable, and maintainable. But deploying AI comes with unique infrastructure and operational challenges—so much so that an estimated 75% of ML models never make it to production due to deployment, monitoring, and management hurdles. So, what does this mean for you as a software engineer? Just as cloud computing became a fundamental skill over the past decade, AI concepts are becoming an essential part of the job. You don’t need to be an AI researcher, but understanding how AI-powered systems work, how they’re deployed, and how to work with them effectively will set you apart. 🚀 To future-proof your career, here are some key areas to focus on: * Start with the fundamentals * Get comfortable with data * Get involved in your company’s existing AI efforts * Get comfortable with AI-assisted coding tools 💡 START WITH THE FUNDAMENTALS You don’t need a data science degree, but you should understand the core concepts behind AI models. What is a linear regression model? How does gradient descent work? What happens when a model trains? There’s a lot of math, theory, and statistics behind these concepts, and it’s hard to build the right intuition just by reading the code. Learn the fundamentals first, then revisit the code—it will make much more sense. Once you have the basics, you can dive into the fundamental mechanics of your specific domain. For example: LLMs, computer vision, or classic ML. This foundational knowledge is especially important for software engineers because AI models behave differently from traditional software. Instead of following strict, rule-based logic, models make predictions based on patterns in data—which means their outputs can vary. Wrapping your head around this difference will make it much easier to design, debug, and work with AI-powered systems. 📚 Resources: Andrew Ng’s Machine Learning Specialization [https://www.coursera.org/specializations/machine-learning-introduction] on Coursera, and Designing Machine Learning Systems [https://www.oreilly.com/library/view/designing-machine-learning/9781098107956/] by Chip Huyen. 📊 GET COMFORTABLE WITH DATA If you’re a software engineer, you’re already familiar with debugging broken systems. But with AI, data is often the biggest source of failure—not the code itself. The definition of “working” becomes much less clear. A model could be functionally working, but still broken for any number of reasons unrelated to the code, such as changes in input distributions, missing values, or inconsistencies between training and production data. These can silently degrade performance over time, which is why observability and data monitoring tools are becoming essential. Being effective in AI requires shifting your mindset from a deterministic lens to a probabilistic one. Exploring datasets yourself, plotting and analyzing distributions, and comparing training to production data are some small ways to get more comfortable here. Without this mindset shift, the non-deterministic nature of AI can easily be a source of frustration if you’re not ready for it. 🤝 GET INVOLVED IN YOUR COMPANY’S EXISTING AI EFFORTS Chances are, your company wants to incorporate AI somehow. If possible, get involved in those efforts. You may not have AI expertise, but you do have software expertise. It’s relatively easy to build an AI proof-of-concept in a Jupyter notebook. On the other hand, it’s hard to deploy production-grade models and do this in a repeatable way that scales as your AI initiatives grow. This is where MLOps (Machine Learning Operations) comes in. Much like DevOps revolutionized the software deployment lifecycle, MLOps applies similar principles to the AI lifecycle–focusing on automation, reproducibility, and monitoring. By adopting MLOps practices, teams can iterate faster, deploy with more confidence, and reduce the operational burden required to keep these models functioning at scale. This is where you can shine as a software engineer. Pretty much everything besides the core model development is an area where you already have the skills to meaningfully contribute, probably more than you think! 🛠️ GET COMFORTABLE WITH AI-ASSISTED CODING TOOLS AI-powered coding tools have come a long way in the last year, and are becoming an integral part of modern engineering workflows, much like version control and automated testing. Tools like GitHub Copilot, Codeium, and Tabnine can help speed up development, but they also introduce new considerations—understanding how AI-generated code interacts with existing architectures, ensuring security best practices, and verifying correctness remain the engineer’s responsibility. If your company allows you to use an AI-assisted IDE, start experimenting. If not, try them out in side projects. Even better, use them to help you learn AI fundamentals—ask them to walk you through building a simple model step by step. Be sure you actually understand what’s happening under the hood, as knowing when (and when not) to trust AI-generated code will be a valuable skill moving forward. Avoiding these tools altogether would be like an accountant avoiding Excel—technically possible, but unnecessarily limiting in an increasingly AI-driven industry. 🚀 CONCLUSION Ignore the doomsayers—software engineering isn’t going anywhere, but it is changing. As AI becomes a standard part of modern systems, the demand for engineers who can bridge the gap between research and production will only grow. The best way to future-proof your career: stay curious, keep learning, and evolve with the industry as it changes. ## Job of the Week [https://go.mlops.community/48cnx9](https://go.mlops.community/48cnx9) ## Hidden Gems [https://go.mlops.community/MLConfess](https://go.mlops.community/MLConfess) Join our next MLOps Community Reading Group to discuss Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations [https://go.mlops.community/zmofln]. Our expert moderators will break down the key findings and what they mean for AI in the workplace, followed by an open discussion where you can ask questions, share thoughts, and challenge ideas. Sign up here [https://go.mlops.community/vm2voe] [Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations](https://go.mlops.community/zmofln) ## TECH TEAZER QUESTION Yesterday was St. Patrick’s Day, and it also marked the start of Nvidia GTC 2025. Founder and CEO Jensen Huang is set to give his keynote today from 10:00 AM - 12:00 PM PDT. If you followed the official Guinness pouring technique, how many perfect pints could you pour during his speech? a) 80 b) 170 c) 60 d) 210 Click here [https://go.mlops.community/TechTeaserAnswers] for the answer ## HERE TO HELP Before you go, here are three ways I can help - just hit reply: * Curated intros to other community members * What problems are you dealing with? Let me help you find the best solutions through my network * Looking to augment your staff for an MLOps or AI project? I got you covered Thanks for reading, catch you next time! Interested in partnering with us? Get in touch: partners@mlops.community Thanks for reading. See you in Slack [https://go.mlops.community/slack], YouTube [https://www.youtube.com/channel/UCG6qpjVnBTTT8wLGBygANOQ?view_as=subscriber], and podcast [https://home.mlops.community/public/content/] land. Oh yeah, and we are also on X [https://twitter.com/mlopscommunity] and LinkedIn [https://go.mlops.community/linkedin]. The MLOps Community newsletter is edited by Jessica Rudd [https://www.linkedin.com/in/jmrudd/]. Shoutout to Ranuga Disansa [https://go.mlops.community/i83caq] for his contributions. --- Source: https://aaif.live/newsletters/mlopscommunity/2025-03-18-want-to-support-the-swes-you-work-with-this-should-help