An independent AI lab, built the long way. We research machine learning and deep learning, and build the products that put it to work.

A paper plane gliding across a soft dawn sky
QUAD CHAT↗70+ MODELS—IMAZHO↘ANNOT8—SPEECH & TTS↗ALBANIAN AI—WYOMING·PRISHTINA·LONDON—QUAD CHAT↗70+ MODELS—IMAZHO↘ANNOT8—SPEECH & TTS↗ALBANIAN AI—WYOMING·PRISHTINA·LONDON—QUAD CHAT↗70+ MODELS—IMAZHO↘ANNOT8—SPEECH & TTS↗ALBANIAN AI—WYOMING·PRISHTINA·LONDON—QUAD CHAT↗70+ MODELS—IMAZHO↘ANNOT8—SPEECH & TTS↗ALBANIAN AI—WYOMING·PRISHTINA·LONDON—

0 +

MODELS IN QUAD CHAT

0

PRODUCTS & APPS

0

RESEARCH AREAS

0

CITIES: WYOMING, PRISHTINA, LONDON

WHY WE EXIST

Most AI software is built fast and carelessly — impressive in a demo, unreliable in daily use. We thought the opposite approach would win, so we built a lab around it.

A single clay vase with one olive branch on a plaster ledge

Research, design and engineering sit at the same table here. A few principles shape every decision.

01

Quality compounds

We'd rather ship one great thing than five mediocre ones — in products, in code and in reputation.

02

One table

The people who design our software are in the same conversation as the people who build and train it.

03

No black boxes

With our users or our partners. You'll always know where things stand, what's working and what isn't.

04

Impact, not output

Success is software people return to every day, not the number of screens or lines of code shipped.

Machine learning and deep learning — from core methods to models for language, voice and vision.

Quad Chat came first. Each product since has been run as a real product — operated, supported and improved continuously.

20252026NEXT

QUAD CHAT LAUNCHES

Every frontier model in one conversation — ChatGPT, Claude, Gemini, Grok and 70+ others under a single subscription.

IMAZHO & ANNOT8

Two new products from the same lab: photos remembered with vision models, and a better way to give feedback on the web.

OUR OWN MODELS

Speech, language and Albanian models of our own — and evaluation methods we intend to share.

20252026NEXT

A deliberately small team working on hard problems — speech, language, safety and Albanian AI. We hire carefully and not on a fixed schedule.

  • Research01

    Machine learning and deep learning research — speech and TTS, language models, evaluation, safety and efficient ML.

  • ML engineering02

    Training pipelines, data tooling, evaluation infrastructure and the systems that turn experiments into models.

  • Product engineering03

    Frontend, backend and infrastructure for products used by thousands of people every day.

  • Design04

    Product and interaction design for AI — including what good voice and multilingual experiences should feel like.

  • Albanian language05

    Native Albanian speakers and linguists who want to shape how AI understands and speaks the language.

  • TELL US WHAT YOU’D WORK ON

The thinking, experimentation and care behind what we research and ship.

01 RESEARCH

02 EVALUATION

03 PRODUCT CRAFT

04 SAFETY

05 THE LAB

A product idea, a partnership, press, or early access to something we’re building. We read everything, and a real person replies within 1–2 business days.

LET’S TALK
Friendly robot with glowing eyes