Selected work

The data and AI chapter at Microsoft, 2020 to 2026, with the two earlier stories recruiters ask about. Each case says what the situation was, what I did, and what came of it; outcomes stay within what colleagues have said publicly.

Download resume ↓

2020 - 2026

Windows quality, measured from preview channels to retail

Situation: Windows ships to more than a billion devices. Engineers needed to know whether a release was healthy before it reached customers, and to find the issues that mattered once it had.
What I did: Led the PM team that built the data products and the review rhythms measuring Windows quality across the Insider preview channels and at retail scale, weaving customer signals and machine learning into insights engineers act on.
Outcome: Data-supported release and quality decisions across hardware, software and services, and faster detection of the issues affecting customers. Colleagues describe these as the critical data products that measured and elevated Windows quality.

As described by Amit Gupte in his recommendation

2024 - 2026

Experimentation and AI for how Windows is built

Situation: As Group PM for Windows & Devices Data, I set the direction for how experimentation, machine learning and AI are used internally to accelerate Windows product development.
What I did: Met weekly with engineering leadership on experimentation and AI problems, and delivered large AI projects myself alongside the team’s roadmap.
Outcome: “Steven completed some large AI projects that even after his departure are having significant impact here.”

As described by Ryan Parsell in his recommendation

2024 - 2026

Forecasting how a Windows AI feature would land

Situation: A key Windows AI feature was heading to retail, and leadership wanted to know how customers would receive it.
What I did: Read the preview-channel signals and forecast the feature’s retail reception well before broad rollout.
Outcome: The forecast held. Amit Gupte calls it the clearest demonstration of “an intuitive understanding of customer behavior second to none.”

As described by Amit Gupte in his recommendation

2024 - 2026

An OKR tool, rebuilt with AI when the old one retired

Situation: The organization’s OKR management tool was being retired, leaving the planning rhythm without a home.
What I did: With a member of my team, ideated and stood up a working replacement from scratch using AI coding tools, rather than waiting for a procurement cycle.
Outcome: The gap was closed with a functional replacement, and the project became the team’s working example of a modern AI development lifecycle.

As described by Amit Gupte in his recommendation

2026

wilssens.com: a family platform built in 16 days

Situation: After Microsoft, I wanted to stay sharp on what platform teams ship by shipping one: email and websites at the family domain for everyone named Wilssens.
What I did: Designed and built the platform with Claude Code: rules enforced by the database, experiments shipped dark and measured, AI help restricted by architecture rather than by prompt, and payments built and switched off.
Outcome: Live and serving the family, with the design decisions written up for hiring managers.

Read the write-up →

2016 - 2019

Spatial audio, from a Windows feature to a platform

Situation: Immersive audio existed in cinemas and on a few headsets. Windows and Xbox needed a platform that developers, Dolby, DTS and streaming services would build on.
What I did: Led the Microsoft Spatial Audio Platform across Windows and Xbox: Windows Sonic, the Dolby Access and DTS Sound Unbound launches, and the partnerships with Disney+ and Netflix that brought Dolby Atmos to both.
Outcome: Monthly active devices playing spatial audio grew more than 250% in 18 months, and the work produced 19 granted US patents.

See the patents →