Product design & AI engineering
I'm Stephen Bowman, a product designer and AI engineer near Chicago. I've spent ten years working on software that helps people make sense of dense information and get things done.
At IBM iX, I worked across aviation, public healthcare, pharmaceuticals, and commerce. My clients included United Airlines, All Nippon Airways, BP, AT&T, Google, California DHCS, and Janssen.
That work included food and coffee ordering prototypes for BP, website research and workshops in Tokyo for All Nippon Airways, wireless customer journeys for AT&T, and a customer-care dashboard for Google. My IBM work also included project coordination for Entergy's smart-meter program.
As an AI Engineer at AstraZeneca, I gathered requirements, built HTML prototypes, and pushed design branches to GitHub for incorporation. The AI workflows explored how people inspect sources, correct inputs, and review generated content.
I research the task, work through requirements, and design the interface in detail. For client teams, working prototypes make behavior concrete enough to question and refine with product, engineering, and domain experts.
Independently, I develop and maintain data applications and AI tools. I use coding agents to explore and implement ideas, then review the changes and check the behavior. CFB Model Lab brings that practice together around a question I care about: what can the numbers tell me about a football team? Building the data processing as well as the interface helps me understand which comparisons I can offer and where I need to explain a limit.
Before product design, I was a digital marketing intern at Grainger and an associate account executive at Yelp. I researched competitors, wrote marketing recommendations, and presented advertising products to prospective customers.
I'm from the Midwest and a new dad. Building with AI keeps me curious. So does figuring out fatherhood.
The work behind the words. Open a project to see the decisions, the original artifacts, and my contribution.
At United, I interviewed and observed customer care representatives in Chicago and Houston. That research helped shape a dashboard around the complaint they needed to resolve across thirteen systems.
Across four DHCS workstreams, I rebuilt Bootstrap patterns in Figma and documented the service rules behind them, so teams could review a shared pattern in the context of each workflow.
For Janssen, I worked within a shared Figma system across eighteen brands. Enrollment steps and exception states made the interaction details available for review alongside the visual design.
In CFB Model Lab, I design the research flow and help build the software behind it with coding agents. Python prepares the data for a Next.js interface. Dated JSON snapshots, the files passed between them, keep the source context available for review.
At AstraZeneca, I used a source-review prototype to onboard designers and worked with product partners to resolve differences between versions. The handoff distinguished working interactions, simulated inputs, and services still to connect.
I build and debug software with coding agents. Here are two contributions accepted by the projects I work with.
Hermes Agent Accepted commit · 4d47891
Dashboard actions in Hermes Agent could inherit another profile’s credentials and settings. I contributed a Python fix that gives named profiles their own environment, with regression coverage.
Read the Hermes fixtoken-tracker Merged pull request · #18
GPT-4o could inherit GPT-4o-mini’s price in token-tracker. My merged fix matches the exact model ID, with regression tests for either order of the returned models.
Inspect the pricing fix