Live a life, man (Day 39)
Hi, it’s a fresh Saturday, and I’m here leaving a message on the internet.
I’ve been focusing on Kubernetes lately. Yeah, on-prem Kubernetes with an ML inference server on it. I wish I could use Triton or something similar, but the situation is that the model is highly customized and I still need to run it on an Nvidia TX2. Yeah, this is what I do for a living.
So here’s what I’ve done: setting up GPU pods, inference servers, module servers, and pipeline servers running with Prefect. My goal was a bit too ambitious—I should’ve focused on batch inference at first. I was trying to handle 70,000 individual requests in about 2–3 minutes, which was insane.
So why am I saying all this? Because I didn’t have much time to work on my game lately. Still, I’ve been pushing the concept a bit further.
Since it’s about talking to NPCs and either initiating trade or waiting for trade, I wanted to bring out more of that merchant vibe. You roam, talk, and start a little selling stand if you wish.
So tomorrow, I’ll focus more on the dialogue system. And if that gets done, I’ll start working on the main game loop.
Thank you for reading.
P.S. Another reason I didn’t make much game dev progress is… I spent some time playing games. Football Manager! It’s a fantastic game, and it’s so satisfying to bring a very low-tier team all the way to winning the EPL. But yeah—I deleted it. It was consuming way too much time.