Building AI systems, fine-tuned language models, and the visual systems that make them legible — bridging the gap between deep-tech and design since 2012.
I'm an AI developer and model post-trainer at TensorVizion with a background that most ML engineers don't have: a decade-plus in graphic design and SEO marketing running in parallel.
That means I build fine-tuned language models with Qwen3 and Qwen2 architectures, ship production-ready UI in Next.js, and understand how to get any of it found on the internet — all at once.
My formal foundation is in Audio Engineering (Honours, Metalworks Institute), which trained my ear for detail, precision, and the difference between a signal and noise — useful in every field I've worked in since.
A bold, dark-mode blog UI built with Next.js and v0, deployed on Vercel. A brutalist visual system designed for developers who want their writing to hit as hard as their code — obsidian palette, strong type hierarchy, and zero visual noise.
A Qwen3-architecture causal language model post-trained and fine-tuned by TensorVizion. Built with Unsloth for efficient training, supports text-generation-inference and conversational endpoints. Apache 2.0 licensed.
A 3.17 billion parameter Qwen2-architecture model quantized to 4-bit with bitsandbytes for efficient local inference. Jinja brings serious model weight to resource-constrained environments without sacrificing capability.
A rigorous technical and creative program in audio engineering, acoustics, signal processing, and studio production. Graduated with Honours — the precision and systems thinking that audio engineering demands translates directly into how I approach AI model training, interface design, and technical optimization today.
Open to collaborations across AI development, UI/UX design, and SEO consulting. If you've got a problem that sits at the edge of those disciplines, I'm probably the right person to talk to.