Microsoft AI Website
A site-wide update to microsoft.ai, expanded and shipped for Microsoft Build.
The Role:
Producer and DRI, with hands-on design and tooling.
Building a Hillclimbing Machine: Launching seven new MAI Models
The challenge:
A site-wide update had to land in time for Microsoft Build. New and expanded model pages, a restructured nav, eight technical blogs, and fresh content across the site, on a hard deadline and without dedicated design support for everything coming in.
What I did:
I owned the project end to end, from kickoff through launch, directing an external design agency and a development agency and aligning ten model PMs across six product categories.
Where I stepped in:
When the resourcing ran short, I filled the gaps myself. I designed and laid out all technical blogs in Figma, using a Gemini workflow to draft each in MAI’s style, then finishing by hand. I recorded the voiceover for the transcription model page when there was no time to hire an agency. And with model data landing last minute, I built a self-serve tool that let PMs enter their own numbers and export drop-ready JSON straight to dev, so the charts populated themselves.
My voiceover
Vibe-coded app
The outcome:
A brand new homepage, 7 model pages, and 9 technical blogs, shipped on time at Microsoft Build, the front door for Microsoft AI on its highest-visibility day of the year.
The Role:
Producer and DRI, with hands-on design and tooling.
Building a Hillclimbing Machine: Launching seven new MAI Models
The challenge:
A site-wide update had to land in time for Microsoft Build. New and expanded model pages, a restructured nav, eight technical blogs, and fresh content across the site, on a hard deadline and without dedicated design support for everything coming in.
What I did:
I owned the project end to end, from kickoff through launch, directing an external design agency and a development agency and aligning ten model PMs across six product categories.
Where I stepped in:
When the resourcing ran short, I filled the gaps myself. I designed and laid out all technical blogs in Figma, using a Gemini workflow to draft each in MAI’s style, then finishing by hand. I recorded the voiceover for the transcription model page when there was no time to hire an agency. And with model data landing last minute, I built a self-serve tool that let PMs enter their own numbers and export drop-ready JSON straight to dev, so the charts populated themselves.
My voiceoverVibe-coded appThe outcome:
A brand new homepage, 7 model pages, and 9 technical blogs, shipped on time at Microsoft Build, the front door for Microsoft AI on its highest-visibility day of the year.