Case study · Construction technology
Proving a faster way to build a fab.
My work at Hoffman spans two periods: keeping the 3D models coordinated on Intel's D1X fab, and now helping teams see how fabs and data centers can be built faster. Both come down to making a complicated process clear enough for people to act on.
Fab D1X, Hillsboro, Oregon. Photo: Hoffman Construction Company
Then · The problem
A coordination process built for ten files met thousands.
D1X was the first fab where everything inside, from mechanical and electrical systems to process piping and tool installation, was modeled in 3D before construction. Under Intel's Copy Exact approach, this fab became the template for Fab 42 and later expansions, so there was no room for rework.
Our coordination process was built for about ten files per building. Trade partners emailed their models daily, and we combined them and checked for clashes. At D1X a single building ran to thousands of files, some of them gigabytes. Email was out, downloads took hours, and nobody could keep track by hand of who had sent what.
Then · The platform
A system that knew what every model needed.
As product manager, I designed a platform where trade partners uploaded their models. It knew every file each building needed, tracked revisions, pulled the latest version of each, and put together the current model for every building, filtered by floor or scope. It also showed at a glance who was and wasn't submitting.
It started in Excel, which broke, then moved to Access, which also broke. We brought in a developer to build a proper database with a simple web front end tied into Hoffman's systems. It handled more than 8,000 files a week across six buildings and over a hundred designers, and carried forward to Fab 42.
Now · The problem
On a fab, speed is worth paying for.
Fabs and data centers start earning money the day they come online, so finishing months sooner is worth a great deal. Getting there usually means changing how the building goes together. Sometimes that's bringing a trade partner on early, which pulls their costs forward. Sometimes it's setting up a temporary factory on site to build large assemblies and lift them into place whole, which adds engineering, temporary works, and cost.
Either way, we're asking a client to spend more now to finish sooner, and their first question is always how. With this many moving parts, nobody can hold the whole plan in their head.
Now · The approach
Sequencing the whole site before anyone breaks ground.
I build simulations that show the construction sequence and the site logistics together: where cranes, deliveries, and crews move, whether a large pre-built module actually fits its path, and how one concrete pour changes the space available for everything else. We use them to test alternatives and to show a client exactly how a faster plan would work.
The sequences are often too complex to animate by hand, and the schedule data is often incomplete, so I use Claude to help program them, which lets us try alternatives quickly. I'm also helping bring AI into the company's work more broadly. The part I enjoy most is making information easy to understand, so the people deciding can see what they're committing to.