PLATFORM

The engine underneath the suite.

One twin engine, one capability record, one validation package, one security posture. This is the part your IT, quality and security teams will ask about, answered in the detail they need.

THE TWIN ENGINE

A machine-exact twin is a manufactured product,
not a bespoke project.

The difference between a generic simulation and one that qualifies an operator is that the second is built to the make and model of the equipment that operator will actually touch. That depth is normally bought at the price of time: a bespoke build per machine, quoted in months, priced as a project.

The twin engine exists to remove that trade-off. Authoring is a pipeline with defined inputs, defined stages and a defined output, which is why depth and speed stop being opposites.

The Inputs

  • Approved SOPs and work instructions
  • Equipment manuals and manufacturer documentation
  • CAD models and P&ID specifications where available
  • Room and area classification
  • Photographs and site walkthrough capture where drawings do not exist

The 5-Stage Pipeline

1. Ingest: Documents and drawings are parsed into a structured step model: sequence, parameters, acceptance criteria, failure modes.
2. Model: The equipment, the room and the interaction surfaces are built to the make and model. Physical accuracy matters because a twin that behaves wrongly teaches the wrong behaviour.
3. Author: Steps are bound to the model, with guided prompts, failure branches, and the acceptance criteria the assessment will enforce.
4. Validate: The twin is reviewed against the source document by your subject matter experts before release, and the review is recorded.
5. Publish: One asset, delivered to the relevant surfaces: headset, glasses, tablet, desktop.
PIPELINE OUTPUT & COMPATIBILITY

What the Pipeline Produces

Because the model is built once and bound to steps rather than baked into a video, the same asset serves simulation, guided execution and, where relevant, robot training. It is authored OpenUSD-compatible so it can be consumed by Isaac Sim and Cosmos without being rebuilt.

PROVED AT SCALE & AUTHORING TIMELINE

More than 150 GMP simulations built to this pipeline, across parenteral and oral solid dosage operations.

Typical machine-specific twin authored in 6–10 weeks, subject to SOP maturity and drawing availability.

CAPABILITY DATA LAYER

One record of
who can do what.

Most plants can tell you who attended a training session. Very few can tell you whether that person can perform the task, which step they fail on, or whether the capability they demonstrated eight months ago still holds. The records exist, but they sit in separate systems and none of them was built to answer a capability question.

The capability data layer is the part of 8chili that makes this an operating system rather than a set of applications. Every interaction across the platform, whether a simulation attempt, a spoken inspection answer, an executed work instruction or a troubleshooting query, resolves to one person, one role, one line, one record.

What It Holds

  • • Step-level telemetry: what was performed, sequence, correctness
  • • Attempt histories, including failures where learning occurs
  • • Competency scores against enforced acceptance criteria
  • • Readiness ratings by role and by theme
  • • Timestamped and attributable floor execution records
  • • Skill decay over time before degrading into deviations

Metrics Produced

DRX (Deployment Readiness Index)Whether a specific person is qualified to be on a specific line today.
DRIFT (Do It Right First Time)Whether that capability is translating into correct execution on the floor.
Skill Drift TrackingTriggers re-qualification based on evidence rather than arbitrary calendar dates.

Architecture Property

Why this is not reproducible by a stack of tools: Telemetry of this kind can only be generated by a system that watches the work being performed against a known-correct model. A learning system records attendance because attendance is all it can observe.

Where it goes: Records flow to your LMS/LRS over xAPI and to your quality system for CAPA linkage.
AI GOVERNANCE

Bounded, logged,
reviewable.

Artificial intelligence in a GxP environment is a validation problem before it is a capability problem. The question is not what the model can do; it is what it is prevented from doing, and whether that prevention can be demonstrated.

1

Grounding

Where 8chili answers a question about a controlled document, the answer comes only from your approved document set. Nothing from the open internet, nothing from another customer, nothing from model memory.

2

Refusal by Architecture

Retrieval runs first. If retrieval returns nothing relevant, the language model is never invoked and the system states that the answer is not in the approved documents. The refusal is a design control in the retrieval layer, not an instruction to the model.

3

Citation

Every answer carries the document, section and page it came from, so a person verifies against the controlled document in seconds. The system points at the source; it does not replace it.

4

Bounded Generation

Conversational interactions run inside an authored scenario frame. The model does not free-run outside it. Every interaction is logged and reviewable.

Version Safety

Superseded revisions are excluded from the index and cannot surface. Re-indexing on document revision is part of the operating model, with a defined trigger and a named owner agreed at deployment.

Human in the Loop

No 8chili system approves a document, releases a batch, closes a deviation or qualifies a person. It produces evidence and answers; your procedures and your people make the decisions.

Regulatory Alignment

This risk-based, design-controlled approach is consistent with the direction of the FDA's 2025 framework on AI credibility in regulatory decision-making and the draft EU GMP Annex 22.

VALIDATION

A documented package,
written for your quality team to review.

8chili ships a computerised system validation package with every deployment, following GAMP 5 and a V-model methodology. It is written to be read by a quality reviewer rather than by a procurement portal.

What the Package Contains

User requirement specification (URS)
Functional specification (FS)
Design specification (DS)
Risk assessment and FMEA with RPN scoring
Installation qualification (IQ) protocol and report
Operational qualification (OQ) protocol and report
Performance qualification (PQ) protocol and report
Requirements traceability matrix (RTM)
Validation summary report (VSR)

What Is Ours and What Is Yours

8chili supplies the package, the protocols and the evidence of system behaviour. Validation itself is site-specific and remains yours: your quality unit executes and approves the qualification for your environment, your equipment and your procedures. The package exists so that execution is a review rather than a build.

Frameworks Aligned To

21 CFR Part 11EU GMP Annex 11GAMP 5ALCOA+ICH Q10EU GMP Chapter 221 CFR 211.25EU GMP Annex 1 (2022)

Tamper-Evident Session Records: Electronic records are attributable, legible, contemporaneous, original and accurate, with tamper-evident audit trails and electronic signature support aligned to 21 CFR Part 11.

Request the Validation Summary
SECURITY

Certified, and reviewable.

ISO 27001

Certified Information Security Management System

SOC 2 Type II

Certified Security, Availability & Integrity Auditing

Security Controls

Encryption in transit (TLS 1.3) and at rest (AES-256)
Role-based access control (RBAC) with granular permissions
Enterprise single sign-on over SAML 2.0
Automated user provisioning and deprovisioning
Audit logging of administrative and data access events
Penetration testing and vulnerability management on a defined cadence
Documented incident response and breach notification

For your security review: Certification reports, architecture documentation and completed security questionnaires are available under NDA.

YOUR DATA

Your content and your records
are yours.

Stated plainly, because it is the question behind most of the others.

Ownership

Your SOPs, your authored content, your operator records and your execution evidence belong to you, without qualification. Termination does not change that; export is contractual.

Controlled Documents

Where 8chili answers questions from your approved documents, those documents remain in your infrastructure. The 8chili layer manages access and usage; it does not hold your document content.

Configurable Capture

Where the platform captures evidence of work being performed, what is captured, at which steps, and for how long it is retained are your configuration decisions, disclosed to your workforce.

Anonymised Aggregate Data

Any use of anonymised aggregate interaction data to improve platform models is explicit, contractual and agreed in advance. It is never assumed, and it is never a default.

Multi-Tenant Cloud Platform & Data Residency

Hosted by 8chili on AWS multi-tenant cloud infrastructure with isolated customer data partitions and regional residency options (including AWS ap-south-1). High availability, continuous security monitoring, and zero infrastructure overhead.

INTEGRATIONS

It fits the estate
you already have.

8chili does not ask you to replace a system you have already validated. It integrates at the layer where workforce capability is created and recorded, and it exports into the systems that already hold your records.

Open Standard Integration Interfaces

8chili Core is engineered with open, standard integration protocols. We seamlessly connect to your existing LMS, QMS, LRS, or MES as your deployment requirements dictate—with zero legacy vendor lock-in.

xAPI 1.0.3 & SCORMStep telemetry and score sync to any conformant LRS/LMS
SAML 2.0 & Okta SSOAutomated identity management, user provisioning & Single Sign-On
REST APIs & WebhooksReal-time execution records, audit archives & QMS CAPA triggers
Standard Audit ExportInspection-ready PDF/CSV session summaries for QMS archives

Third-party names are the trademarks of their respective owners. Their appearance indicates interoperability and not partnership or endorsement.

DEPLOYMENT

Multi-tenant cloud architecture,
built for scale.

Hosted by 8chili (AWS Cloud)

Multi-tenant cloud platform hosted on AWS with regional data residency options (including AWS ap-south-1). Fastest to start, fully managed, and zero customer IT hardware footprint required.

Multi-Device Ecosystem

Standalone VR headsets, with no PC tether and no dedicated room. AR smart glasses specified for use in Grade A and B environments. Tablets, smartphones and shop-floor terminals. Device management and deployment guidance is included in implementation.

SCOPE

The things we
deliberately do not build.

A platform that claims to do everything is a platform that will be evaluated against everything. 8chili occupies one layer and integrates with the rest.

We do not build an MES

Manufacturing execution is a validated system you already own. 8chili sits above it in the workforce execution layer. No rip and replace, no platform war, and a much shorter path to a pilot.

We do not build a document management system

Your controlled documents stay in your DMS under your change control. 8chili reads the approved set; it does not author, revise or approve content.

We do not replace your LMS

Your training record stays where it is. 8chili supplies better evidence to put in it.

We do not make the decision

No 8chili system releases a batch, closes a deviation, approves a document or qualifies a person. It produces the evidence; your procedures and your people decide.

GET STARTED

Bring us your hardest question.

A technical review with your IT, quality and security teams: architecture, validation package, security posture, integration scope.