CR3ATE.AI

This is where things started getting genuinely interesting.

CR3ATE.AI sat at the intersection of computer vision, AR/VR, and real-time ML, and my time there was less of a gradual ramp-up and more of a straight jump into the deep end. Which, honestly, was exactly what I needed.

A significant chunk of my work was on CoreML model development for iOS, specifically around real-time blur detection in camera applications. That meant evaluating and optimising architectures like MobileNetV3 and EfficientNet in TensorFlow, benchmarking them against each other, and figuring out what actually held up under the constraints of a mobile device.

On-device inference is a different beast from running things on a server, and getting that right requires you to care about things like latency and model size in a way that most ML work simply doesn't force you to.

The other thread I was pulling on was integrating Hive APIs to support 3D Gaussian splat reconstructions for AR, VR, and mixed reality workflows. If you haven't come across Gaussian splatting yet, it's one of the more exciting developments in 3D scene reconstruction and the applications in spatial computing are pretty significant. Getting that pipeline to work end-to-end was one of the more technically satisfying things I've done.

On the backend, I was working with custom NestJS APIs, containerised via Docker and wired up to MongoDB to support the ML-driven functionality sitting on top of it. I also spent time on AWS, writing Lambda functions for file processing and handling uploads to S3, which fed directly into the broader processing pipeline.

Beyond the individual pieces, I co-built the production iOS app and the customer-facing web frontend and backend for cr3ate.ai itself, connecting everything to the ML pipeline through the NestJS APIs and AWS infrastructure. End-to-end, not just one layer of the stack. That included Firebase authentication, JSON parsing across services, and keeping the whole thing coherent as the product was actively being shipped.

This was the first time I was working on something that was live, customer-facing, and evolving in real-time. The ML work was interesting, but shipping production software on top of it taught me something different altogether.

The overlap between applied ML and full-stack infrastructure is messy in the best possible way. CR3ATE.AI is where I started to get comfortable operating in that overlap.

These guys are genuinely doing some insanely cool work and before I signed off from here, we'd begun setting up the foundation for many agentic workflows too building entire automation systems that could fire off with just a prompt and self healing loops for times when a reconstruction doesn't come out just right (artists are finicky, what can I say).

Sony is their client and they demo-ed our pipeline at CES 2025, not to mention they've worked with Netflix, Amazon and a lot of government contractors and large scale VFX studios as well.

Oh and did I mention we got our app (the one that I co-built) published and live on Apple's AppStore as well? Yep that was an insane achievement in itself if you know how harrowing Apple's approval process is 🙄.

Check out what these guys do at cr3ate.ai and my official certificate (my badge of honor) here.