Vincent Wendy

Vincent Wendy brings a product designer’s approach to the visual work of large technical conferences. Credited as AI Engineer’s senior creative designer, he described producing landing pages, speaker announcements, stickers, track mascots, signage, and digital room schedules as the team’s only designer. His contribution extends beyond individual graphics: he builds reusable layouts and AI-assisted workflows that help a small team produce consistent materials, check for omissions, and respond to changes.

A product design foundation for conference materials

Wendy’s earlier background in product design informs how he approaches an event. He uses atomic design to break layouts into small, reusable pieces, then combine them into complete deliverables. He also thinks through the attendee’s journey: finding registration, reading a map, following wayfinding, scanning a QR code, and reaching the right room. These materials need to work together because each helps someone take the next step.

In One Designer + AI. Hundreds of Deliverables., Wendy explains how that approach changes conference production. Hundreds of speakers and a large sponsor roster create repeated work with consequential details: a missing sponsor logo or an incorrect room schedule can make a polished graphic fail its purpose. He starts by defining typography, colors, components, and spacing. Explicit desktop and mobile type sizes give agents concrete instructions instead of leaving them to improvise. Once the website establishes those rules, marketing colleagues can use the same visual foundation for emails, flyers, and documents.

AI-assisted workflows for recurring production

His AI-assisted design workflow combines Figma, GPT, and the coding agent Devin, with requests and feedback moving through Slack. Figma specifications provide measurements for spacing, font sizes, and colors; Model Context Protocol connections can supply additional context. Wendy uses these inputs to make the intended layout easier to implement and correct. For his conference work, he describes a shorter cycle from a request in Slack to a Figma design, back to the agent for implementation and revision.

• Room schedules: Room schedules show how this works in practice. Layouts previously assembled manually in Figma can be generated from the latest conference data for a selected room and day, then exported as PNGs for on-site screens. The workflow still needed adjustment: one morning, a schedule change exposed the absence of an editing control. Wendy asked Devin to add an edit button, changed the schedule, and exported a replacement image. Handling exceptions became part of the production tool rather than a separate manual task.
• Speaker announcements: Speaker announcements use the same principle of defining a layout once and making its variations repeatable. Wendy demonstrated a generator that combines names, headshots, and speaker details in portrait and landscape formats, alongside trading-card designs. The tool makes individual graphics available without requiring him to assemble every announcement separately. Shared rules preserve the visual pattern while the underlying details change.
• Visual quality checks: He also uses agents to assist with visual quality checks. For sponsor banners and shirts, he asks Devin to compare the graphic with the expected logos and flag omissions, supplementing human inspection. He reported that his own sponsor-logo tests found every omission; that result describes his tests, rather than establishing a general accuracy guarantee. In a separate demonstration, he showed a photograph-matching interface that lets him review whether candidate images depict the same speaker, using Jason Liu as the example. The practical benefit is finding a usable speaker photograph without searching a photographer’s files one by one.

Working around model limitations

Wendy treats a model’s limitations as a reason to change the workflow. Using a pelican riding a bicycle as an illustration example, he described direct vector generation producing an unusable result. His alternative was to generate a raster image and vectorize it in Figma. The useful question is how to obtain a deliverable with the tools available, rather than whether one model can perform every step.

His position is that concrete production problems give designers direction when AI makes execution easier. Repeated assets can be automated, but the designer still needs to identify what can go wrong and make corrections practical. Wendy’s conference work joins those responsibilities: establish the visual system, make recurring work repeatable, inspect the output, and keep the attendee’s experience in view.