December 19, 2022. The air is crisp, the days are shorter, and my thoughts wander to the tech world’s flurry of activity over the past year. AI/LLM infrastructure was the buzzword du jour; I’ve been deep in that space with projects leveraging large language models for various applications. Meanwhile, platform engineering was becoming a recognized discipline within teams, with everyone from ops to dev starting to see its value.

One of my biggest takeaways this year is just how much FinOps and cloud cost pressure are shaping the way we build systems. As an engineer, it’s hard not to get excited about new tech, but when you’re staring at a monthly bill that can grow quickly, those considerations start to weigh heavy on the design process.

I recently worked on a project that involved building a serverless architecture using AWS Lambda and API Gateway. It was a perfect opportunity to try out some of the latest tools in the ecosystem. Serverless is an area where the tech stack is always evolving—new languages, new frameworks, and continuous improvements in performance and cost efficiency.

The journey began with fine-tuning our models for a specific use case. We were leveraging Riffusion for music generation, which is a fascinating space given how much creativity it opens up. Integrating this into our serverless architecture required some careful thought about cold starts, memory management, and overall performance optimization. After all, you don’t want to blow through your Lambda budget with unexpected costs!

One of the biggest challenges was ensuring that our music generation tool didn’t hit its concurrency limits. We had to do a lot of load testing and tweaking of function configurations to get it right. This experience really hammered home just how critical it is to have robust observability and logging in place. Without these, you’re flying blind.

Speaking of monitoring, another area I’ve been diving into more deeply this year is metrics and cost tracking. The DORA (DevOps Research and Assessment) metrics are increasingly influencing my approach to platform engineering. We started implementing automated cost reports and alerting for our services. It’s not just about saving money; it’s also about ensuring that we can scale resources appropriately without overprovisioning.

Looking at the Hacker News stories, the fusion news was mind-blowing. Imagine harnessing the power of a star in your data center—almost! But I couldn’t help but think about how this tech might one day influence edge computing or even cloud infrastructure design. It’s crazy to imagine.

On the flip side, the Twitter account drama just reminded me that even the biggest platforms can have their moments. And John Carmack’s departure from Meta was a bit of a shocker; his work has always been innovative and influential. His move to a smaller company might signal some interesting shifts in the tech world’s direction.

As we wrap up 2022, it feels like we’ve come full circle with serverless technology. Back in 2016, it was all about “serverless” hype, and now here we are—still using functions as a service, but with a much deeper understanding of the trade-offs involved. It’s exciting to see where this will go next year.

And so, as I close my laptop and wrap up another busy day, I find myself thinking about what 2023 might bring. Will LLMs continue to evolve? How will FinOps impact our architecture decisions? And what new serverless tools will make their way into the toolbox?

Merry Christmas to all, and here’s hoping for a tech-filled New Year!