Dark theme
WELCOME TO QQUANTUM

Let Qubi show you around.

Choose how you’d like to explore.

Don’t agreeGo to yourOS instead ↗

Both choices allow Google Analytics, so we can see which pages help.

Sound is your choice. Essential cookies keep the site working. Continuing either way also allows Google Analytics (visits and pages viewed).

Read our Terms, Privacy and Cookies.

Change your choices anytime from the footer.

Getting Qubi ready.
Reference Architectures · 5 pages

AI reference architectures.

Published designs for the systems we build most often — the components, the failure modes, and the numbers you should hold a vendor to. Written to be useful whether or not you hire us.

Introduction transcript. Client: What’s actually underneath an AI system? Qubi: Let’s lift the lid. Carefully. Qubi: Data, models, evaluation, and the connections between them. Client: And the bits that could go wrong? Qubi: Those too. A useful design shows its failure modes. Qubi: Explore the plans. Ask better questions.

At a glance

Checkable, not unfalsifiable.

01 The problem
An agency case study reports a percentage and a happy client. The percentage cannot be verified, the architecture is not shown, and it is unclear whether the result transfers.
02 What we publish
The component graph with the models named, the failure modes, and the evaluation that tells you the design works. An engineer reading it can tell if it is wrong.
03 The figures
None of the figures here are claimed client outcomes. They are design targets and published benchmarks, labelled as such.
04 Use
Written to be useful whether or not you hire us. Each one lists what to measure before you believe anyone, including us.
WHY THESE EXIST

A design an engineer can argue with.

A reference architecture is the opposite trade to a case study. It shows the actual component graph, names the models and infrastructure, states where the design breaks down, and lists what you should measure before you believe anyone — including us.

These are designs we work from, documented in the abstract. They are not client reports, and none of the figures on these pages are claimed client outcomes — they are design targets and published benchmarks, labelled as such.

How to read these
Problem
What breaks, and why the obvious fix does not work
Design
The component graph, with the models named
Failure modes
Where it degrades, and the mitigation
What to measure
The evaluation that tells you it works
THE POSITION

The standard we hold ourselves to.

A percentage you cannot verify is marketing shaped like evidence. A design an engineer can argue with is the real thing.
PICK ONE

Four architectures, documented.

Pick the system closest to yours. Each one shows its components, where it breaks, and what to measure.

Pick one

Design targets, not results from a specific deployment. The steps are taken from each architecture's own page.

Healthcare · Legal · Insurance

Document intelligence over regulated archives

Extracting structured fields from tens of thousands of long, unstructured PDFs without sending a page to a public endpoint — and making every extracted value open to the exact sentence it came from.

Sectors
Healthcare · Legal · Insurance
Design
Air-gapped inference, constrained decoding, span-level citation
  1. Long PDFs
    Unstructured archives, no public endpoint
  2. Air-gapped inference
    Constrained decoding, structured fields out
  3. Span-level citation
    Each value to its exact sentence
  1. 1Failure: scanned pages older than the OCR. Gate each page on OCR confidence before extraction runs.
  2. 2Failure: tables that span pages. Reattach the header, and range-check every numeric field.
  3. 3Failure: near-verbatim spans. Normalise whitespace and ligatures only, never content words.
  4. 4Failure: a conservative threshold sends everything to review. Tune it against a labelled set.
  5. 5Measure: field-level accuracy against a labelled holdout, reported per field.
  6. 6Measure: review rate, the share of documents routed to a human.
  7. 7Measure: cost per document, fully loaded with GPU time, storage and review minutes.
SaaS · E-commerce

Support deflection that knows when to stop

A retrieval agent over product docs and account state that answers what it can evidence and hands off everything else — with the handoff threshold as an explicit, tunable parameter rather than an emergent accident.

Sectors
SaaS · E-commerce
Handoff
An explicit, tunable threshold
  1. Hybrid retrieval
    Over product docs and account state
  2. Confidence gating
    Answers only what it can evidence
  3. Human handoff
    Everything it cannot evidence
  1. 1Failure: documentation that is quietly out of date. Add freshness metadata, and exclude sources past a staleness threshold.
  2. 2Failure: multi-turn drift. Re-classify scope on every turn, not only the first.
  3. 3Failure: users who learn which phrasing gets a human. Watch the escalation-reason distribution over time.
  4. 4Measure: deflection at a 7-day reopen window. A ticket that reopens was not deflected.
  5. 5Measure: answer accuracy on a weekly sample of deflected tickets, graded against the human QA rubric.
  6. 6Measure: escalations with usable context attached, and unsupported claims reaching a customer.
Financial services

Risk and compliance agents with an audit trail

Agents that read policy, apply it to a case, and produce a decision a regulator can reconstruct — every tool call, retrieved passage and intermediate judgement persisted as an immutable record.

Sectors
Financial services
Record
Every tool call, passage and judgement persisted
  1. Deterministic tools
    Policy applied to the case
  2. Four-eyes review
    A second reviewer on the decision
  3. Full trace
    Immutable, step by step
  1. 1Failure: model deprecation. Know the window before you commit, and keep an open-weight fallback you can host.
  2. 2Failure: rationales that read well and explain nothing. Require a cited policy clause ID, and reject any without one.
  3. 3Failure: reviewers who rubber-stamp. An override rate at or near zero means the control is not real.
  4. 4Failure: snapshot drift. Run tools with no network, so a live read fails loudly in development.
  5. 5Measure: replay fidelity. Re-run archived traces; the identical-decision rate should be 100%.
  6. 6Measure: rationale citation coverage, the share of decisions citing at least one policy clause.
Logistics · Supply chain

High-throughput document pipelines

Multimodal extraction at volume, where the interesting problem is not accuracy on a good scan but throughput economics and what happens to the small fraction of documents that are damaged, rotated or in an unexpected language.

Sectors
Logistics · Supply chain
Economics
Cost per document, at volume
  1. Batch inference
    Multimodal documents, throughput economics
  2. Exception routing
    Damaged, rotated, unexpected language
  3. Cost per document
    What the pipeline costs at volume
  1. 1Failure: template drift. Track accuracy per template and per sender, and alert on a per-sender drop.
  2. 2Failure: exception queues that grow without bound. Treat queue depth as a production SLO with an alert.
  3. 3Failure: batch latency colliding with operations. Run two lanes, synchronous for time-critical documents and batch for the rest.
  4. 4Failure: costs that only look good per document. Measure cost per successfully emitted record.
  5. 5Measure: straight-through rate, the share of documents reaching the ERP with no human touch.
  6. 6Measure: exception mix over time. Whichever type dominates is what to fix next.
EVERY ARCHITECTURE PAGE

All four, as pages.

Hire your first
AI employee.

Tell us the job you would hand over if you trusted a machine with it. In one 30-minute call we will tell you whether it should, what it would take — and if it should not, we will say so.

Pick the job — the form arrives ready for it
Book a 30-minute call How we work
30-minute technical call · no deck
Shipped & in build

Real work. Check it yourself.

What we built, what stage it is at, and a link to the live thing wherever there is one.

All work →
The updated yourOS homepage: Roxy’s open ring surrounded by bento cards for calls, voice, translation, building, design, apps and more.
The yourOS homepage on a phone, with its compact navigation, welcome film button and bento cards.
youros.tech — the homepage bento hero
Our own · coming soon

yourOS

yourOS is an operating system with an AI at its heart. You talk to it like a person, out loud or typed. It makes your calls, plans your days and works on your projects with a crew of agents, on your own private server. Our own product, built from the first commit; the waitlist is open.

What we built
  • Operating system
  • Roxy
  • Agent crew
  • Voice calls
  • App store
  • Website
Meet yourOS →
The Sonora homepage on desktop: a full-bleed dune landscape behind the headline “Transform Your Life with Sound”, with App Store and Google Play download buttons.
The Sonora homepage on a mobile viewport, showing the same hero reflowed to a single column.
sonora.com — homepage, 1440×900
Live · iOS · Android

Sonora

Website, backend and app assets for Sonora, a free AI sound-therapy app on iOS and Android published by Cognitive AI.

What we built
  • Website
  • Backend
  • App assets
Read the case study →
The updated Podswap homepage on desktop: “Grow on Instagram and TikTok for free, with real engagement from creators your size”, beside a tilted wall of exchange, tool and guide cards on a violet gradient.
The updated Podswap homepage on a phone, with its floating navigation, Instagram and TikTok engagement headline and tilted tool cards.
podswap.app — homepage, 1440×900
Five platforms

Podswap

Brand, website, prototype, apps on five platforms, the AI backend and the algorithm-intelligence pipeline for Podswap, a social growth engine that runs without a human in the loop.

What we built
  • Brand
  • Website
  • Prototype
  • Apps on five platforms
  • AI backend
  • Algorithm pipeline
Read the case study →
The unOpen.AI homepage introducing Roxy, a private AI assistant that remembers who you are.
The unOpen.AI homepage on a mobile viewport.
unopen.ai — homepage, 1440×900
Prototype · closed beta

unOpen.AI

In closed beta: Roxy, a private AI assistant with permanent memory running on infrastructure the user owns. Brand, site, product, backend and conversational design.

What we built
  • Brand
  • Site
  • Product
  • Backend
  • Conversational design
Read the case study →
coherence.com on desktop, headed "Intelligence that moves in phase with people".
coherence.com on a mobile viewport, reflowed to a single column.
coherence.com — homepage, 1440×900
Live

Coherence

Live at coherence.com: brand identity generated with AI agents we directed, the motion system and website for the group behind Sonora, AboutAI, WakeUp and MudShadows. Agent work remains in build.

What we built
  • Brand identity
  • Motion system
  • Live website
  • Agent work — in build
Read the case study →
The new Domain Booth homepage on desktop: the headline “Premium Domains. Serious Buyers. Proven Results.” beside the animated DB monogram.
The new Domain Booth homepage on a phone, reflowed to a single column.
Our build of the Domain Booth homepage — 1440×900
Live

Domain Booth

Live at domainbooth.com for DomainBooth, a premium domain brokerage: a new animated mark, every page redesigned and rebuilt with designed mobile layouts, and a custom CMS for its sales record, featured names, press and enquiries.

What we built
  • Animated mark
  • Every page redesigned
  • Designed mobile layouts
  • Custom CMS
Read the case study →
Who we have built for

Twenty-one years. Names you know.

Apps, platforms and campaigns for these teams, long before the AI work. Every one is written up on the work page — what we did, and where our part ended.

See all of our work →
  • BBC Good Food logo
  • Charities Aid Foundation logo
  • GAME logo
  • NatWest logo
  • Next logo
  • NSPCC logo
  • Tesco logo
  • Worldpay logo
  • Samsung logo
  • Sony logo
  • Land Rover logo
  • UAP Limited logo
  • Netflix logo
  • Amazon Prime Video logo
  • Mercedes-Benz dealerships logo
  • San Miguel logo
  • Biovene Barcelona logo
  • Seventy7 logo
  • Littlewoods logo
  • Smile logo
  • The Co-operative Bank logo