Amazon AWS
Six years defining UX for generative AI, computer vision, and edge computing — translating highly technical AWS systems into clear, customer-centered experiences at scale.
- AI/ML Product Design
- Enterprise SaaS
- Developer Tools
- Design Vision & Direction
- Computer Vision
Six years at Amazon AWS defining and driving UX vision for generative AI, computer vision, and edge computing services — aligning product, engineering, and executive stakeholders around a unified experience strategy in ambiguous, engineering-driven environments.
Selected screens
Amazon Bedrock — Generative AI Platform · 2023 – Present
Promoted to L6 during my Panorama tenure, I now drive UX vision for the next generation of Amazon Bedrock, AWS’s fully managed generative AI service serving 10,000+ enterprise customers.
Bedrock service vision (current)
Leading the design direction for the next major revision of the Bedrock console — defining the end-to-end UX strategy for how enterprise customers discover, build, and scale AI applications. Driving alignment across product, engineering, and design through iterative vision artifacts and structured team critique.
Prompt Management & Caching — flagship
Primary UX designer for Prompt Management, enabling enterprises to create, version, and deploy prompts at scale across foundation models. I untangled layered complexity between model-provider capabilities and platform-level controls to simplify how customers define cacheable prompt elements (system instructions, tools, messages).
- Launched Prompt Management Caching in preview at re:Invent 2024; GA April 2025.
- Up to 90% cost reduction and 85% latency reduction through cached reuse.
- Care Access: 86% lower data-processing costs, 66% faster processing, 4–8+ hours saved daily.
Guardrails — account & organization enforcement
Synthesized enterprise security-team feedback to simplify enforcing AI safety policies at scale — collapsing guardrail policies, cross-region inference definitions, and account-level enforcement into a single-definition model for consistent safeguards across many accounts and applications.
Flows — multi-turn conversation support
Designed agent-node functionality inside a visual AI workflow builder, letting developers build dynamic conversational applications without complex coding, with executive-level sign-off. Thomson Reuters and Dentsu Creative cited the “intuitive visual interface” and “complex, flexible, multi-prompt workflows.”
Navigation & overview
Volunteered to consolidate navigation and information architecture across all Bedrock design contributors ahead of re:Invent, shipping a unified service overview organized around the customer’s AI product-building journey.
AWS Panorama — Computer Vision at the Edge · 2020 – 2023
AWS Panorama is an edge ML appliance and SDK that brings computer vision to existing on-premises cameras.
Service launch
Owned the redesign of resource-management screens, moderated research sessions with ProServe and Solution Architects, applied a holistic CFS matrix to measure action severity across the console, and resolved 200+ UX bugs to reach launch readiness.
Defect Detection
Designed the console workflows that let customers build and deploy computer vision defect-detection apps using Amazon Lookout for Vision — without ML expertise. Tyson Foods automated inspection of 8,000+ product carriers per line, hitting 99.1% anomaly-detection accuracy and saving an estimated 15,000 hours of skilled labor annually.
Hardware ordering & quoting
Removed a critical procurement friction point — customers had waited days for a quote via a third-party email process. Designed an integrated quote-and-purchase flow using an external vendor’s API with real-time shipping cost, reducing quote generation to under one minute.
Remote troubleshooting suite
After SSH access was removed at GA, designed troubleshooting within strict security constraints: remote reboot (resolved ~80% of customer issues), enhanced CloudWatch log deep-linking, and encrypted device-log transfer over USB with LED signaling. Deloitte/VicDot and TaskWatch confirmed it addressed their primary operational pain points.