SIEMORA — AI-native simulation experience platform. Learn, simulate, practise, decide, master.

Anyone can watch a video.
Prove they can do the work.

SIEMORA is an AI-native platform where a learner performs the real task, not a multiple-choice version of it. Every action is captured, scored against rules your institution sets, and kept as replayable evidence — so a competency claim can still be defended years after the certificate was printed.

01
LEARN
Video, reading and micro-video bound to the moment
02
SIMULATE
The real material, in the browser
03
PRACTISE
Guided, then unaided, as often as needed
04
DECIDE
Judgement under the conditions of the job
05
MASTER
Competency backed by evidence, not attendance

The gap

Content is nearly free now. Proof is not.

AI made producing courses cheap, and every institution now has more material than its learners will ever watch. What no institution can buy off the shelf is a defensible answer to the only question an employer, a regulator or a parent actually asks:

“Can this person actually do it — and how do you know?”

Completion percentages do not answer it. Quiz scores do not answer it. A recorded assessment answers it once, for one assessor, and then decays into a video nobody will ever scrub through again.

What most platforms measure

Minutes watched. Modules completed. Questions answered correctly. All of it is a proxy for attention, not for capability — and all of it is now trivially gamed.

What SIEMORA measures

What the learner did, in what order, how precisely, how independently, how they recovered from an error, and whether they checked their own work before submitting.

And keeps

An append-only record of every attempt, stamped with the exact specification and rubric version that scored it. Replayable action by action, for as long as the record exists.

How it works

An experience, not a playlist.

Your institution composes what each topic requires — stage by stage. A theory topic might be video and a quiz. A practical topic runs the full pipeline. A judgement topic branches into consequences. The platform executes whatever you compose; it never dictates a single teaching method.

YOUR CONTENT PLATFORM-OWNED SIMULATION YOUR STANDARD Video & reading your material Micro-video bound to a timecode Simulation the real task Guided → independent hints on, then off Assessment your rubric Competency with evidence RECORDED THROUGHOUT every selection · every placement and its precision · every hesitation · every hint opened · every rule passed or failed every self-correction · every retry · the spec, rubric and content versions in force at that moment

You compose the pipeline

Choose which stages a topic uses and in what order. Four presets to start from, or build your own. Validation refuses to publish a pipeline that cannot be completed or scored.

You bring the content

Your curriculum, your videos, your readings, your quizzes, your rubrics, your assessors, your certificate. Nothing about your method is imposed.

We own the simulation

What happens inside a simulation — the rules, the state machine, the critical errors, the scoring — is the platform's, so a result means the same thing everywhere it is issued.

Everything is versioned

Publishing a new specification never rewrites a record already issued. An attempt keeps the exact versions that scored it, permanently.

The simulation engine

Built domain-neutral, on purpose.

The engine knows about materials, states, rules, errors and evidence. It knows nothing about any particular profession — and a build check fails if a domain-specific term ever appears inside it. Each field is added as a domain pack on top.

That is why a second discipline costs content and rules rather than a rebuild, and why a learner on one device sees the same object, scored the same way, as a learner anywhere else.

1

The material behaves like the material

Real dimensions, real colours, real tolerances. A step that is wrong is not rejected with a buzzer — it is left visibly wrong, so the learner sees the error before anyone tells them.

2

Scoring is server-authoritative

The same rules run in the browser for instant feedback and again on the server as the authority. A result cannot be produced by a modified client.

3

Practice and assessment are one engine

Hints, retries, difficulty and guidance are configuration, not separate builds. The task a learner rehearses is the task they are assessed on.

4

It works when the network does not

Practice, hints and the full event log run offline on the device and reconcile when the connection returns. Nothing is lost, nothing is duplicated.

Live example · a sequencing task

Ordering ten graded rods shortest to longest, edges aligned. The uneven step is left in place deliberately.

Step 5 jumps by three units instead of one. The engine records rule.fail ascending_order, the coach asks the learner to look along the right-hand edge, and the correction they make themselves is scored higher than never having erred.

What one attempt records

Events captured162
Rules evaluated6
Placement precision0.9 mm
Replayableindefinitely

AI-native

AI that watches the work, not the transcript.

Most “AI learning” products generate content and answer questions about it. SIEMORA's intelligence sits on a structured stream of what a learner physically did — hundreds of typed events per attempt, with precision, timing, ordering and recovery. That is a far richer signal than text, and it is what makes the following possible rather than decorative.

While the learner works

A coach that knows what just happened

Feedback is generated from the event stream and the rule trace — the specific step, the specific deviation, the specific pattern across attempts.

  • Speaks to the error actually made, not a generic tip
  • Withholds the answer when the method requires discovery
  • Adapts tone as frustration or fluency shows in the timings
  • Reads aloud, for learners more comfortable listening
After the attempt

Explainable scoring

Every result opens into the reasoning behind it: which rule decided it, at which second, and what the learner would need to do differently.

  • Rule-by-rule trace linked to the moment it fired
  • Plain-language explanation of a numeric score
  • Comparison against the learner's own previous attempts
  • No black box — the score comes from rules, the words from AI
Immediately after

Remediation bound to the second

When a rule fails, the platform does not say “watch the video again”. It opens the exact segment that teaches that step and returns the learner to the mat where they left it.

  • Micro-content indexed to topic, stage and timecode
  • Selection driven by the failure pattern, not the topic
  • Effectiveness measured — did it change the next attempt?
  • Repeated ineffective remediation flags a content gap, not a learner gap
For the trainer

Triage across hundreds of learners

No trainer can inspect five hundred learners. The platform ranks who needs a human today, and why, with the evidence attached.

  • Critical, attention and normal, from event patterns
  • Distinguishes a method error from a pace problem
  • Flags disengagement before it becomes withdrawal
  • Every recommendation carries the attempt behind it
For the author

Assisted authoring

Turning a syllabus into a working course is the slowest part of any programme. AI drafts; your experts approve.

  • Draft objectives, scenarios and question banks from source material
  • Suggest a pipeline shape for each topic and flag unscoreable ones
  • Transcribe, chapter and index every video automatically
  • Propose micro-video cut points from the transcript and the rubric
Across the cohort

Competency inference and integrity

Patterns that no individual attempt reveals: where a whole cohort is weak, and where a result does not look like honest work.

  • Weakest competency surfaced across cohorts and trainers
  • Topics generating the most errors, ranked
  • Anomalous attempt patterns raised for human review
  • Predicts non-completion early enough to act on it

The line we do not cross

AI explains a result. It never decides one.

A competency record has to be reproducible from the event log and the rules that were in force — by a regulator, an employer, or a court, years later. A model's opinion is not reproducible, so scoring never depends on one. AI narrates, recommends, triages and drafts; rules score, humans decide, and every override is recorded.

Model-independent records

The model can be replaced without invalidating a single record already issued, because no record depends on it.

Metered and capped

AI spend is metered per institution against a ceiling. At the cap, features degrade to authored fallbacks — they do not fail.

Your data stays yours

Each institution runs on its own isolated database. Aggregate improvement is permitted; seeing another institution's learners is not.

Evidence & competency

A certificate says they attended. A record says what they can do.

SIEMORA keeps four things separate, because conflating them is the fastest way to a claim an institution cannot defend.

ConceptWhat it meansWho determines it
CompletionThe learner worked through the contentPlatform, from progress records
CompetencyThe learner performed to a rubric, with evidence, repeatedlyPlatform, from the evidence chain
CertificateA document issued on the institution's own criteriaYour institution
AccreditationWhether an external body or employer accepts that certificateNeither of us alone

What SIEMORA can defend is the second row — and it is the row that becomes more valuable, not less, as AI-generated credentials multiply.

Append-only, never edited

Events are written once. A correction is a new event, not an overwrite, so the history of an attempt cannot be quietly improved after the fact.

Replay from the log, not a recording

The attempt is reconstructed from the events themselves, so it replays identically on any device, at any moment in the attempt, for as long as the record exists.

Four levels, not a pass mark

Competency moves through defined levels on accumulated evidence, with thresholds, expiry and reassessment your institution sets.

Portable when you want it

A signed, verifiable credential a learner can present outside the platform — so what they earned is theirs, not locked to a login.

For institutions

Your method. Our machinery.

The boundary is written down and does not move. It is what lets an institution teach its own way while a competency record still means the same thing everywhere it is issued.

You own

  • The curriculum, the syllabus, the topics and their objectives
  • Which stages each topic uses, and in what order
  • Your videos, micro-videos, readings, documents and images
  • Your quizzes, assignments and assessment rubrics
  • Your live classes, on whichever provider you already use
  • Your assessors, their judgement, and their right to override
  • Your certificate, its wording and its awarding authority
  • Your learners' data, in your own isolated database

SIEMORA owns

  • What happens inside a simulation — rules, states, critical errors
  • Scoring, server-authoritative and reproducible
  • The event log, its integrity and its versioning
  • The competency model and its evidence chain
  • The AI layer, its metering and its guardrails
  • Multi-tenant isolation, proved automatically on every release
  • Accessibility, offline operation and device support
  • Uptime, backups and disaster recovery

Small institution

One cohort, one trainer, a laptop and a phone. Composed pipelines, hosted content, live classes on the tool you already pay for, and certificates issued under your own name.

Multi-centre

Branches, sections and departments, with cohort-level and organisation-level competency reporting, and trainers who see only their own learners.

Enterprise & regulated

Single sign-on, directory sync, retention policy, audit trail, API access, and the evidence chain that compliance training is judged on.

Platform

Engineered for records that outlive the software.

97
modules specified before a line was written
1,160
API endpoints, documented and contract-tested
8
architecture checks that fail the build, not a review
4
device widths designed, phone to desk

Isolation, proved

A database per institution. No query, job, cache key or channel can cross a tenant boundary, and an automated prover checks it on every release rather than a policy asserting it.

Accessible by construction

Every material interaction completes by tap as well as by drag, every control is a real control, and every coach message exists as text and as speech.

Offline-first practice

A learner on an unreliable connection keeps working. Events queue on the device and reconcile without duplication when the network returns.

Open standards

An xAPI-shaped event log, SCORM import for existing material, Open Badges export for credentials, and a documented API for everything the interface can do.

Multi-tenant SaaS Database per institution Server-authoritative scoring Append-only event log Versioned specifications Offline PWA SSO & directory sync SCORM import Open Badges Full REST API Audit trail Disaster recovery drills

Where it goes

One engine. Many disciplines.

The first domain pack is early-years teacher training, delivered as a one-year diploma with a partner academy. It was chosen because the method is demanding, the materials are physical, and success is judged on how a teacher behaves in front of a child — a genuinely hard test of the engine.

The architecture was built so the second discipline costs content and rules, not a rebuild. Any field where someone must demonstrate a procedure, under conditions, to a standard, is the same problem.

Teacher training Clinical & allied health skills Laboratory & QC procedure Industrial safety & compliance Equipment operation Hospitality & service standards Field service Regulated financial process

First programme

DurationOne year, online
Topics180
Practical simulations77
Competencies24
First cohort[intake]

Partner academy, programme fee and cohort size available under NDA.

See a learner build the thing, fail, and be corrected by the material.

A walkthrough takes about thirty minutes. We will show you a real simulation, the attempt it produces, and the replay an assessor sees a year later — using your discipline if you send us one topic in advance.

Available on request: the platform specification, the module and API inventory, and an interactive prototype you can operate yourself.