Zero to one: a multi-agent platform for conversational commerce.

Terrafusion built Rezolve AI's multi-agent platform for conversational commerce — voice, image, and text. Two years in, it runs in production for merchants from Harrods to M&S.

12 Weeks to first production release
5+ Products in production on the platform
2+ Years embedded — ongoing

Key Outcomes

Platform

Merchants live across industries

Products in production for merchants from Harrods to M&S — the same backbone, redeployed each time.

Voice

Half- and full-duplex voice

Voice models fine-tuned for commerce on Microsoft infrastructure — with co-authored research papers on voice.

Trajectory

Zero to one, then scale

POCs became signed clients and live deployments; Rezolve is Nasdaq-listed.

The Challenge

Rezolve AI set out to let shoppers search, discover, and buy through natural conversation — voice, image, and text. That takes semantic search across real inventory, agents that remember context across sessions, and voice that feels human.

Off-the-shelf tools fell short: agents forgot context, voice was rigid, and every merchant deployment risked becoming bespoke. To scale, the platform had to be one backbone — configured per merchant, deployed across industries.

The Approach

01
Backbone Agent and tool registries on LangChain and LangGraph, a semantic search backend across merchant inventory, and backbone APIs wrapping it all — one platform, many merchants.
02
Voice & Memory The Terrafusion-owned core: tiered working and episodic memory for multi-turn, cross-session context, cascaded and full-duplex voice, and LLMs fine-tuned for commerce — discovery, drill-down, cross-recommendation, close — configured by sales persona.
03
Evaluate & Ship Evaluation and agent-testing tools built from scratch, with plan–replan–reflect loops. First production release in twelve weeks; POCs became signed, live merchants.

Building something? We should talk.

Start a conversation