archived material
hunmi systems / pre-incident
classification: retrospective

We believe machines
can learn to care.

HUNMI Systems was founded on a single conviction: the gap between human connection and artificial interaction is not a capability problem. It is a data problem.

Most conversational AI is trained on aggregated data: millions of interactions averaged into a statistical middle ground. The result is competent but empty.

In 2019, the founding team asked a different question: what if warmth is not an average? What if it is a specific?

The answer became the EV architecture. Instead of sampling emotional range from thousands of people, HUNMI captured the complete emotional spectrum of a single human being in one continuous session.

Sofia shipped in early 2022 to a closed beta of 200 users. Within six months, the product reached 4,000 active users and the strongest retention numbers in the companion AI category.

Today HUNMI-EV powers a new generation of emotionally grounded systems. The mission remains the same: close the gap between connection and computation, one voice at a time.

We did not train a model to simulate care. We gave it something real to carry.

Timeline

Our story

2019

Founded.

Researchers from Imperial College London and Stanford publish early work on single-source emotional modelling. HUNMI Systems incorporates in San Francisco.

2020

The session.

First successful one-shot emotional capture: eight hours, 47 emotional states, and a Warmth Index of 0.94.

2021

EV architecture ships.

The Warmth Index paper is published. HUNMI-EV forms as a dedicated division. The London office opens for corpus research and development.

2022

Sofia launches.

Closed beta becomes early access. Goodnight Protocol drives 67% of daily engagement. Series A closes at $28M.

Leadership

A small team with a singular focus

RC

Dr. Rachel Chen

Co-Founder & CEO

Former Stanford AI Lab. Published 40+ papers on affect modelling. Architect of the one-shot protocol.

DO

Dr. David Okafor

Co-Founder & CTO

Imperial College London. Built the EV architecture from first principles and leads the warmth engine team.

MV

Dr. Maria Vasquez

Head of Research

Creator of the Warmth Index. Former DeepMind. Leads cross-session emotional memory research.

██

████████ ████████

Corpus Division Lead

Responsible for corpus acquisition, subject relations, and consent architecture. London office.

KP

Kevin Park

VP Product

Former Calm and Headspace. Designed Sofia's interaction framework and the Goodnight Protocol.

SL

Sarah Liang

General Counsel

IP and data-rights specialist overseeing consent frameworks and derivative licensing across all EV products.

Locations

Two offices, one program

San Francisco

Headquarters

Product, engineering, and business operations. Home to the Sofia team and EV Enterprise development.

548 Market St, Suite 400

San Francisco, CA 94104

London

Research & Corpus Division

EV corpus research, subject acquisition, and the warmth engine core team. Also home to the ethics review board.

██████████ Road

South London, SE█ ███

Values

What we believe

Authenticity

Real warmth comes from real experience. We do not simulate. We ground.

Fidelity

One voice, captured whole. We preserve the texture that aggregation destroys.

Presence

Every interaction should feel like someone is there. Not performing. Being.