Alexa Shopping

Conversational design across agentic commerce, multimodal interfaces, and inclusive AI

Note: All opinions are my own and do not reflect the company or its partners. This is my POV on work I contributed to and I’ve recreated visuals for confidentiality.

Note: All opinions are my own and do not reflect the company or its partners. This is my POV on work I contributed to and I’ve recreated visuals for confidentiality.

Conversational design across agentic commerce, multimodal interfaces, and inclusive AI

Note: All opinions are my own and do not reflect the company or its partners. This is my POV on work I contributed to and I’ve recreated visuals for confidentiality.

Alexa Shopping

Conversational design across agentic commerce, multimodal interfaces, and inclusive AI

Note: All opinions are my own and do not reflect the company or its partners. This is my POV on work I contributed to and I’ve recreated visuals for confidentiality.

Note: All opinions are my own and do not reflect the company or its partners. This is my POV on work I contributed to and I’ve recreated visuals for confidentiality.

01

ALEXA, BUY FOR ME

Monitor deals and purchase on your behalf

Before agentic commerce was a category, I grounded design in familiar mental models, like setting a max bid on eBay. I built the workflows, ran multiple research studies, led guardrail debates, and shipped with engineering. It launched as Amazon's first agent-completed purchase as a flagship goal, and became the foundation Rufus and Alexa+ later built on.

Before agentic commerce was a category, I grounded design in familiar mental models, like setting a max bid on eBay. I built the workflows, ran multiple research studies, led guardrail debates, and shipped with engineering. It launched as Amazon's first agent-completed purchase as a flagship goal, and became the foundation Rufus and Alexa+ later built on.

01

ALEXA, BUY FOR ME

Monitor deals and purchase on your behalf

Before agentic commerce was a category, I grounded design in familiar mental models, like setting a max bid on eBay. I built the workflows, ran multiple research studies, led guardrail debates, and shipped with engineering. It launched as Amazon's first agent-completed purchase as a flagship goal, and became the foundation Rufus and Alexa+ later built on.

02

PRODUCT DETAILS LLM REDESIGN

Static GUI to Dynamic Conversations

1 in 2 customers abandoned the details page on Echo devices, because it couldn't answer the product questions that drive a purchase decision. As one of 3 core designers, I helped rebuild the navigation alongside our science teams to understand our LLM capabilities. This included a modular category-aware layout and a Q&A that synthesized specs, reviews, and seller content.

1 in 2 customers abandoned the details page on Echo devices, because it couldn't answer the product questions that drive a purchase decision. As one of 3 core designers, I helped rebuild the navigation alongside our science teams to understand our LLM capabilities. This included a modular category-aware layout and a Q&A that synthesized specs, reviews, and seller content.

02

PRODUCT DETAILS LLM REDESIGN

Static GUI to Dynamic Conversations

1 in 2 customers abandoned the details page on Echo devices, because it couldn't answer the product questions that drive a purchase decision. As one of 3 core designers, I helped rebuild the navigation alongside our science teams to understand our LLM capabilities. This included a modular category-aware layout and a Q&A that synthesized specs, reviews, and seller content.

03

INCLUSIVE RESEARCH

Expanding Alexa's NLP Training Data

For customers with non-dominant accents, and for low-income households sharing devices on spotty connections, Alexa failed them. More errors, more repeats, more reasons to give up. I co-led a research initiative that mapped these response failures with diary studies, interviews, and literature reviews. For cross-org adoption, I socialized these findings via workshop facilitation.

For customers with non-dominant accents, and for low-income households sharing devices on spotty connections, Alexa failed them. More errors, more repeats, more reasons to give up. I co-led a research initiative that mapped these response failures with diary studies, interviews, and literature reviews. For cross-org adoption, I socialized these findings via workshop facilitation.

03

INCLUSIVE RESEARCH

Expanding Alexa's NLP Training Data

For customers with non-dominant accents, and for low-income households sharing devices on spotty connections, Alexa failed them. More errors, more repeats, more reasons to give up. I co-led a research initiative that mapped these response failures with diary studies, interviews, and literature reviews. For cross-org adoption, I socialized these findings via workshop facilitation.

04

ALEXA, THANK MY DRIVER

Human Connection at Scale

Not every project is about conversion. Sometimes the question is whether a product can make people feel something real. "Alexa, thank my driver" was Amazon's best-performing holiday campaign ever, with 1 million thank-you's in 36 hours. I designed both the holiday and evergreen navigation experiences for future promotions.

Not every project is about conversion. Sometimes the question is whether a product can make people feel something real. "Alexa, thank my driver" was Amazon's best-performing holiday campaign ever, with 1 million thank-you's in 36 hours. I designed both the holiday and evergreen navigation experiences for future promotions.

04

ALEXA, THANK MY DRIVER

Human Connection at Scale

Not every project is about conversion. Sometimes the question is whether a product can make people feel something real. "Alexa, thank my driver" was Amazon's best-performing holiday campaign ever, with 1 million thank-you's in 36 hours. I designed both the holiday and evergreen navigation experiences for future promotions.

Note: All opinions are my own and do not reflect the company or its partners. This is my POV on work I contributed to and I’ve recreated visuals for confidentiality.