Month 5 Week 2: GPU Optimization and Launch Preparation

This week I worked on GPU complexities like installing a second GPU and being able to fit another card into the K8 GPU AI machine. Getting the AI pipeline more performant has been the main focus—it's all about optimizing now and bug bashing to make the whole user experience better as we get closer to launch and demoing the product.

GPU Hardware Challenges

Fitting a second GPU into the Kubernetes GPU AI machine wasn't straightforward. Space was tight, and I had to make sure everything would actually fit and work together. After some trial and error, I got it in there and it's helping with the performance.

Pipeline Optimization

With the extra GPU in place, I've been tweaking the AI pipeline to get better performance. There's a lot of fine-tuning involved—memory allocation, inference parameters, making sure resources are used efficiently. Every bit of optimization helps when you're trying to keep conversations feeling natural.

Bug Bashing Mode

Right now it's mostly about finding and fixing bugs, smoothing out edge cases, and making sure the user experience is solid. The core functionality works, but there are always little things that need attention—how it handles interruptions, silence detection, keeping the conversation flowing naturally.

Launch Prep

I'm working on a presentation video for Milestone 3 since this is the week before the last week of capstone. The journey is about to end and a new beginning is about to happen. It's been a long road from concept to working system, and we're almost there.