About

What I'm doing now

Most of my time goes into Nexus, the internal platform at Tandem Interactive. The agency runs its client work through it, and I designed and built it myself. An agency this size would normally rent that kind of software and then work around whatever it rented. Building our own means the platform fits how the team already works, and the AI inside it is held to a standard we set. I spent ten years as the person this kind of software gets sold to, and that experience shapes what I build into it.

I also build Bobs Studio, the desktop app I do my own generation work in. It bundles a native inference engine that supports more than twenty model architectures across image and video. The GPU work runs in a separate process and streams previews back while it renders. There is usually a checkpoint training in the background, too. Nepotism, the FLUX model I publish, is on its twelfth version.

The job covers more than code. I decide how the company uses AI, including the security around it and the standard anything has to meet before it goes in front of a client.

How I got here

I started freelancing at eighteen and kept it going for about three years. Prodigy Brands took the first real chance on me in 2016. I shot product photography and laid out quarterly catalogs there, and the executives wanted to see drafts before anything shipped. Explaining a design to people who were not designers was good practice for the client presentations that came later.

In 2017, I joined Apple Insurance as a one-person marketing department. I handled the design and copy and built the website, which suited me. The site needed SEO, so I learned it and did it myself, and SEO has been part of every job since. After that I spent eleven months at Southeast Publications, a high-volume shop producing print for parks across the US and Canada. The pace there taught me to design for someone who will look at a piece for two seconds.

Tabanero Hot Sauce brought all of that together in one job. I designed everything the brand put in front of a customer, from the labels and billboards to the sales sheets. I also ran the social accounts, shot the photography for them, and handled the website and SEO. Making every piece myself taught me to plan a campaign as one conversation across every channel.

At Equiturn, I ran the marketing department and its client accounts, and that is where I got serious about measurement. At the peak I had twenty-five accounts running at once, and I took every client meeting myself. Every number I presented had to hold up against people who ran businesses for a living, so I started building reporting that answered the hard question before it was asked. I also trained the new hires and interns there, and I liked that more than I expected to.

Zomee was a single brand, and I ran all of its marketing for about two years. Revenue went from $12M to $25M in eighteen months. The technical SEO rebuild under that growth mattered most for what came next. I did enough of the rebuild myself to learn that I would rather build the thing than write a ticket asking someone else to.

In 2023, I taught myself the machine learning stack so I could train image models of my own. That became Nepotism, and then Tenosai, a company I founded and ran for a year. I shipped its Discord bot, trained its model, and dissolved the LLC when my cofounder and I disagreed about where it was going. What I kept from it was the habit of shipping software on a schedule to people who tell you the moment it breaks.

That background is why the Tandem job fits. I build the platform the agency runs on, and I spent ten years as the kind of person who uses it, so I know which features people open every day and which ones they stop opening by the second week.

How I work

I have presented reports and I have sat through them, and a metric nobody can trace back to a source loses the room fast. Everything I build is designed for the moment someone pushes back on a number. The same goes for the models I publish. I still answer every comment on them myself.

Teaching

Building the platform is about half the job. Teaching people to use it is the other half, so I wrote the course that goes with it. The course runs ten chapters on how these models work underneath. It is all narrated, with a check between chapters so nobody gets carried past something they missed. I run the live sessions myself because the questions people ask tell me what to build next.

Less of it is about the software than I expected. Most sessions come down to sitting with someone who has decided they are bad at this and finding the one idea that is in the way. Usually nobody has explained that idea in plain language before. That was true when I taught reporting dashboards, and it is still true with models.

What I build with

Everything here is something I've shipped in a real project.

Current focus

Essays go up when something I am building is worth writing down. New Studio work goes up when a generation is worth keeping. I also take on select consulting when the problem is interesting.

Get in touch

Frequently asked questions