Selected work
Client and employer details are necessarily selective, but the scope should be clear: multiple full-code products, enterprise agents, retail and retrieval workflows, creative applications, production infrastructure, and the enablement required to make them useful.
Built multiple AI-enabled applications for a global sportswear enterprise, working across product definition, full-stack implementation, managed deployment, iteration, and handoff.
Delivered client-facing and internal conversational experiences designed for high-use retail workflows, alongside retrieval-augmented search and data-processing systems grounded in enterprise information.
Worked directly with design and sportswear teams to apply generative AI to product ideation, variation, visualization, and review—inside the way creative teams actually work.
Built applications and workflows around the Nano Banana family, OpenAI omni image capabilities, and video-generation tooling—translating fast-moving model features into repeatable creative production.
Developed virtual try-on applications and supporting workflows for sportswear use cases, connecting image models, product context, and user experience into an applied system.
Built and supported code-first and low-code agents across Google and Microsoft ecosystems—choosing the delivery pattern around the workflow, controls, and team rather than forcing every use case into one stack.
Helped teams move from access to adoption across GitHub Copilot, Codex, Claude Code, Google AI Studio, Agent Platform, Microsoft 365 Agent Builder, Copilot Studio, and Foundry.
Designed and hardened high-volume AI workflows for generation, transformation, enrichment, and integration—where reliability, recovery, throughput, and cost mattered as much as model output.
Designed and built an in-house orchestration layer with the internal team at a global sportswear enterprise, providing reusable foundations for AI applications whose needs extended beyond a single model call or vendor-native builder.
The common thread
Successful AI work rarely hinges on the model alone. It depends on clean system boundaries, measurable quality, sensible failure modes, and people who know how to operate what was built.
See how I engage →