Hello there,

I’m Thomas, an AI Software Engineer with more than 10 years of experience building AI and software systems.

These days, I work at the intersection of LLM systems, AI accelerators, and production AI: benchmarking distributed inference, studying model architectures and parallelism strategies, working with accelerator software stacks, and building tools to reason about latency, throughput, and system performance.

My path started in computer vision and has taken me through autonomous vehicles, industrial inspection, audio AI, model quantization, AI compilers, NPUs, generative AI, and large language models. The thread has always been the same: I like understanding what happens between a model and the system that runs it.

Over the years, I’ve worked at Evollabs, Synaptics, Roviero, Ignitarium, Infosys, and TrustCircle, while also contributing to projects for organizations including Biotactic, SOF London, Shiba Inu, Sony, Alstom, and others.

What I enjoy most is taking something that works as an experiment and turning it into a dependable production system. Increasingly, I’m also interested in designing systems that other engineers can understand, validate, maintain, and build upon.

Alongside my full-time work, I collaborate with teams around the world on challenging problems, from scientific computer-vision systems to generative AI applications and automation.

A lot of my learning happens simply out of curiosity: experimenting, benchmarking, reading papers, writing small tools, or disappearing into a technical rabbit hole for longer than originally planned.

Outside work, I’m a husband, traveler, occasional gamer, and lifelong learner. Right now, I’m fascinated by Chip War and the history, economics, and geopolitics behind the semiconductor industry. I also love good music and exploring new places.

This site is where I share what I’m learning, projects I’m working on, and ideas I find interesting.

You can find more of my work on GitHub, connect with me on LinkedIn or Twitter, or check out my freelance work on Upwork.

Best,
Thomas