01
Product direction
Turn the companion concept into clear workflows, surfaces, claims, and boundaries.
Case study · Zero to one
Voice AI built as a product, not a demo.
Adrien led Amigo from product vision to a pilot-ready voice AI platform designed for repeated, trusted use.
Snapshot
- Product type
- Voice AI companion platform
- Role
- CTO and co-founder
- Stage
- Concept to pilot-ready product
- Surfaces
- Web, mobile, dashboard, voice layer
Challenge
Make repeated conversations useful, safe, and operable.
The hard part was not making an agent speak. It was building a product people could trust over time.
Conversation flow, latency, memory quality, summaries, operator visibility, and privacy boundaries all affected one another.
That turned every product decision into a system decision: what to remember, what to surface, how to structure the data, and where human review belonged.
Ownership
Product and technical direction stayed connected.
As CTO and co-founder, Adrien took Amigo from concept to a working multi-surface product.
02
Architecture
Design voice sessions, memory, structured outputs, dashboards, and data as one platform.
03
Hands-on engineering
Build across Next.js, TypeScript, PostgreSQL, voice AI providers, and React Native.
04
Pilot readiness
Prepare the system for field use while keeping privacy, tone, and product claims precise.
System
Voice was the interface. Structured context was the product.
The platform separated conversation, memory, summaries, and operator visibility so each layer could evolve without losing coherence.
Conversation system
From conversation to useful operational context.
5 surfaces
01
Voice
The user starts with a natural conversation.
02
Transcript
The call becomes structured text the system can reason from.
03
Memory
Useful context is extracted and kept for the next interaction.
04
Summary
The important signals are turned into a readable update.
05
Dashboard
Teams can review activity, context, and follow-up needs.
Delivered
One product across four connected surfaces.
01
Voice experience
Real-time multilingual conversations with session control, transcription, and synthesis.
02
Memory and data
Persistent context and relational structured outputs designed for repeated use.
03
Web and mobile
User-facing product surfaces built on the same product logic and account model.
04
Team dashboard
Activity, summaries, and follow-up context without making raw transcripts the product.
Outcome
Pilot-ready, without overclaiming what AI should do.
Amigo reached working conditions across voice, memory, structured data, web, mobile, and operator workflows—while maintaining clear boundaries around companionship, care, and diagnosis.
Contact
Bring the product problem.
Share the roadmap, the system, or the decision that has become too important to leave fragmented.