My name is Chris Balmes. I design how AI talks to people.
For nearly a decade I’ve designed voice and multimodal assistants: Amazon Alexa, the Astro home robot, and now Meta AI on smart glasses. I design model behavior, define the evaluations that measure conversational quality, and partner with researchers to turn product taste into specs, data, and evals.
Portfolio available by request — including recent Meta AI work: cbalmes2@gmail.comWhat I do
1. Model behavior
Response principles, ideal responses, and system prompts that shape how a model speaks.
2. Evaluation
Rubrics and LLM judges, calibrated against human raters, that set the quality bar.
3. Real-time voice
Turn-taking, wait-time bridges, earcons, and speech-to-speech delivery.
4. Interaction Design and Prototyping
Interactive voice prototypes with real TTS, used to settle design questions with evidence.
Background
Eighteen years, from editorial analytics to defining how AI speaks.
2011
Oxford University Press
Digital analyst, editorial
2013
Blue Fountain Media
Head of UX design
2015
Smart Design
Senior interaction designer
2017
Amazon Lab126
Senior voice UX designer, Alexa & Astro
2019 Present
Meta
Principal Product Designer, Model Behavior & Voice
Recent Projects in the Press
Shipped work, 2021 — 2025. Case studies are available by request.
Amazon Astro · 2021
Amazon's home robot — Alexa with wheels, expressive eyes, and a periscope camera.
Ray-Ban Meta · 2023
AI glasses with camera, open-ear audio, and Meta AI — smart glasses gone mainstream.
Orion by Meta · 2024
Meta's AR glasses prototype — holograms overlaid on the real world, in glasses form.
Meta Ray-Ban Display · 2025
In-lens display plus neural wristband — messages, captions, and directions at a glance.
Older Case Studies
Selected work in depth, 2016 — 2019. Newer ones are available by request.
What’s on TV? · 2017
Turning 10,000 weekly TV questions — succeeding 7.8% of the time — into a personalized, actionable guide.
Local Search on Echo Spot · 2018
An adaptive UI for finding places by voice — the answer reshapes itself around the question you asked.
Neural TTS · 2019
Emotion, expressed: the research, tenets, and prototyping behind Alexa’s expressive neural voice.
Patents
Three granted, as inventor.
Question answering for a multi-modal system
US 10,929,601 · Granted Feb 23, 2021
Generating context-aware rendering of media contents for assistant systems
US 11,563,706 · Granted Jan 24, 2023
Active listening for assistant systems
EP 4,327,199 · Granted May 6, 2026
Inventor on 11 filed patent applications.