February 2022 felt like a month where the tech world was on fire, both literally and figuratively. The invasion of Ukraine sent shockwaves through our personal lives and professional ones alike, but for us in ops and infrastructure, it brought an unusual calm to our workdays, perhaps because we couldn’t focus much on the war when dealing with serverless chaos.

The LLM Infrastructure Boom

Since ChatGPT’s release, there has been a mad dash to build AI/ML systems that can handle vast amounts of data. Our team was no exception. We spent a lot of time optimizing our infrastructure for handling large language models (LLMs). It felt like every morning I would start with another round of debugging and scaling issues. The excitement of what these models could do overshadowed the daily grind, but there were moments when I wondered if we were pushing too hard.

Platform Engineering Takes Center Stage

Platform engineering became a buzzword this month, as everyone wanted to be part of it. In our company, we had debates about whether platform engineers should focus on just the infrastructure or also take on application development. My opinion? Stick to the infrastructure. I believe that focusing solely on the platform allows us to create reusable components and services that can benefit any developer, regardless of their expertise.

Cost Pressure from FinOps

With cloud costs continuing to rise, our company dove headfirst into FinOps—a discipline aimed at optimizing cloud spending. We started implementing cost controls using tools like AWS Cost Explorer and Datadog. One particularly frustrating day was when we found that a service we thought had been shut down was still running in the background, costing us money unnecessarily. It’s moments like these that make me appreciate the importance of continuous monitoring.

WebAssembly on the Server Side

WebAssembly (WASM) has been gaining traction for server-side applications as well. Our team was exploring how to use WASM to run compute-intensive tasks more efficiently. The initial experiments were promising, but we encountered issues with performance and integration. Debugging WASM was like trying to solve a Rubik’s cube in the dark; every solution seemed to create new problems.

DORA Metrics and Real Work

Despite all the buzz around new technologies, I found myself focusing on old standards. At our weekly standup, we talked about Lead Time Distribution (LTD) and Change Failure Rate (CFR). These metrics are critical for understanding how our engineering teams operate. Last week, a change introduced by one of my colleagues caused a minor outage. It made me realize that even with the latest tools, people still need to understand what they’re doing.

Personal Learning

This month was filled with small wins and setbacks. I learned a lot about serverless architectures, but more importantly, I realized how important it is to keep your feet on the ground—both literally and figuratively. When the pressure of scaling LLMs becomes overwhelming, taking a step back and focusing on what really matters can make all the difference.

Conclusion

February 2022 was a month of contrasts—tech turmoil juxtaposed with real-world events. As we move forward, I hope that our focus remains on creating robust, reliable infrastructure that supports both AI advancements and day-to-day operations. After all, it’s not just about building the next big thing; it’s about building something that stands the test of time.

---

That was my February 2022—full of challenges, but also full of learning. What about yours?