Loading…
Loading…
From Product Specifications to Shipment Release. Every Quality Decision Accounted For.
Introduction
Loretto AI Labs developed PQIMP (Product Quality & Inspection Management Platform), an enterprise-grade quality management solution designed to digitalise, automate and intelligently supervise the entire product inspection lifecycle.
PQIMP addresses one of the most significant challenges facing manufacturers, international sourcing businesses, importers and quality assurance organisations: maintaining consistent product quality across multiple suppliers, production facilities, inspection teams and international supply chains.
PQIMP replaces disconnected spreadsheets, manually prepared reports, scattered photographs, email approvals and inconsistent interpretation of standards with a unified platform connecting product specifications, Standard Operating Procedures, pre-production validation, Golden Sample approval, production inspections, statistical sampling, defect management, corrective actions and shipment authorisation.
Its architecture establishes a complete relationship between what a product is required to be, how it should be manufactured, what inspectors actually observe and whether the finished goods satisfy the agreed quality requirements. PQIMP is more than an inspection application — it is a comprehensive digital quality governance platform for modern manufacturing and global sourcing.
Project at a Glance
The architecture supports organisations managing multiple products, suppliers, factories, inspection teams and customer accounts.
| Component | Platform capability |
|---|---|
| Platform | Product Quality & Inspection Management |
| Core application | Enterprise web portal and inspection workflows |
| Inspection lifecycle | Pre-production through final shipment release |
| Statistical inspection | Controlled AQL sampling |
| Quality framework | ISO 2859-1 methodology, subject to authorised standards access |
| Defect classification | Critical, Major and Minor |
| Product governance | Specifications, SOPs and Golden Samples |
| Evidence management | Photographs, documents and inspection records |
| Corrective actions | NCR, CAPA, rework and reinspection |
| AI architecture | Multi-agent quality intelligence |
| Security | Role-based access, OIDC and PKCE |
| Cloud infrastructure | Railway |
| Identity management | Keycloak |
| Evidence storage | Cloudflare R2 |
| Database | PostgreSQL |
| Platform validation | Approximately 1,647 passing automated tests |
Complete Product Quality Lifecycle Management
Delivered platformQuality managed as a continuous process — not an isolated event at the end of manufacturing.
Every inspection is linked to the approved product specification, relevant SOP, production lot, inspection criteria and quality requirements. Each stage produces traceable records that become part of the product's permanent quality history.
This structured lifecycle helps organisations identify quality problems earlier, improve manufacturing consistency and reduce dependence on fragmented manual processes.
Every stage produces traceable records that become part of the product's permanent quality history.
Stage 1
Product Creation
SKU and product record
Stage 2
Specification Development
Version-controlled requirements
Stage 3
SOP Approval
Checkpoints and criteria
Stage 4
Pre-Production Validation
Risks found early
Stage 5
Golden Sample Approval
Controlled reference
Stage 6
Production Authorisation
Readiness verified
Stage 7
In-Process Inspection
Issues caught mid-run
Stage 8 · Quality gate
Final AQL Inspection
Statistical sampling
Stage 9 · Quality gate
Quality Disposition
Deterministic rules
Stage 10 · Quality gate
Shipment Release
Authorised QA approval
Intelligent Product Specification & SOP Management
Delivered platformA controlled quality baseline against which every manufacturing and inspection activity is evaluated.
Inspection procedures can be configured around individual products, product categories and customer requirements, giving organisations consistent inspection checkpoints, acceptance criteria and evidence requirements.
The objective is that inspectors across different factories and geographic locations evaluate products against the same approved quality expectations.
AI-assisted specification analysis can support the identification of incomplete requirements, ambiguous inspection instructions and potential inconsistencies between product specifications and SOPs.
Pre-Production Validation (PPV)
Delivered platformFormal quality checkpoints during product development and production preparation — not just at the end.
PQIMP's structured Pre-Production Validation process identifies product and manufacturing risks before full-scale production is authorised.
Following successful validation, an approved Golden Sample becomes the controlled reference for subsequent manufacturing and inspection. PQIMP maintains the relationship between that reference sample, the product specification and production inspection requirements — an important safeguard against uncontrolled product variation.
Commercial benefitIdentify and resolve quality issues before they become large-scale production defects.
AQL-Based Final Random Inspection
Delivered platformControlled Acceptance Quality Limit sampling — with acceptance decisions made by deterministic software, not generative AI.
The platform is designed around controlled statistical inspection practices, including the ISO 2859-1 framework, with the applicable standards tables and profiles governed by authorised licensing and configuration.
Statistical acceptance is calculated by deterministic software from an approved, version-controlled AQL profile. The AI layer explains, suggests and flags — it cannot change the result.
Deterministic quality engine
Rules decide · auditable software logic
AI advisory layer
Explains, suggests, flags
01Lot defined
Production lot and quantity recorded against the product
02Approved AQL profile
Version-controlled, configured under authorised standards access
AI advisory layer
Explains requirements
Plain-language explanation of the approved sampling plan
Advises, cannot override
AI advisory input to step 2: advises, cannot override the deterministic result.03Sample size per profile
Determined by the approved profile — never by AI
04Defects recorded
AI advisory layer
Suggests defect class
Proposes a category and severity for the inspector to confirm
Advises, cannot override
AI advisory input to step 4: advises, cannot override the deterministic result.05Accept / Reject numbers evaluated
Defect counts compared with the profile's limits in software logic
AI advisory layer
Flags anomalies
Highlights unusual inspection patterns for QA review
Advises, cannot override
AI advisory input to step 5: advises, cannot override the deterministic result.06Disposition
Accept or reject — repeatable and auditable
Authorised QA approval
Qualified personnel confirm the disposition and any controlled override, with full audit history.
Conceptual only — no ISO 2859-1 table values are shown. Sampling tables are applied from authorised, licensed standards configuration.
PQIMP does not rely on generative AI to determine statistical acceptance or rejection. AQL calculations and mandatory quality rules are processed through deterministic software logic. AI provides supporting analysis, recommendations and anomaly identification without overriding approved inspection standards — combining advanced intelligence with predictable, auditable quality decisions.
Defects presenting unacceptable safety, regulatory or other critical risks.
Significant non-conformities that may materially affect product functionality, usability or agreed quality requirements.
Less significant deviations from approved product specifications.
Digital Inspector Application
Delivered platformA guided inspection workflow that keeps observations, photographs, defects and specifications connected.
Inspectors working in manufacturing facilities and supplier locations are guided through approved inspection checkpoints, with each observation tied to the product specification it is assessed against.
The platform architecture includes offline-capable Android inspection workflows for environments with unreliable factory connectivity. Inspection information is collected locally and synchronised through controlled processes when connectivity returns — particularly valuable for international sourcing operations involving remote manufacturing facilities.
Photographic Evidence & Digital Traceability
Delivered platformPhotographs and documents linked to the inspection, specification, defect and corrective action they support.
PQIMP uses Cloudflare R2 object storage for durable management of inspection evidence and generated documents.
Organisations can investigate historical quality issues without relying on photographs distributed through email, messaging applications or individual inspectors' devices.
Non-Conformance & Corrective Action Management
Delivered platformRecording a defect is only the start. PQIMP carries it through to verified correction.
The platform maintains traceability between the original inspection finding, the corrective action, its subsequent verification and the final quality disposition.
AI-assisted analysis can identify recurring defect patterns and support investigation of potential underlying manufacturing problems.
Each step stays linked to the original inspection finding and its evidence.
Step 1: Inspection Finding
Defect recorded with evidence
Step 2: NCR Raised
Severity and supplier linked
Step 3: Root Cause & CAPA
Corrective action assigned
Step 4: Rework & Reinspection
Correction verified
Step 5: QA Closure
Final disposition recorded
Multi-Agent Artificial Intelligence
Proposed AI capabilityPQIMP provides a foundation for a coordinated network of specialised AI agents designed to support quality professionals throughout the inspection lifecycle. Rather than replacing qualified inspectors or quality managers, the proposed architecture enhances their capabilities by analysing specifications, inspection findings, photographic evidence, historical defects and supplier performance.
The Master Quality AI Supervisor coordinates the specialised agents, consolidates findings, resolves competing recommendations and presents actionable insights to authorised quality personnel — while evaluating agent performance, monitoring recommendation quality and maintaining decision traceability.
Specialist agent network · 19 agents
Operating principle
Agents advise. Deterministic rules decide statistical acceptance, and authorised QA personnel approve.
Selected agent
AI is advisory: it cannot waive safety requirements, alter AQL decisions, bypass approvals or authorise shipment release.
Deterministic Quality Engine
AQL calculations and mandatory quality rules in auditable software logic — AI cannot alter them
Authorised QA Approval
Qualified personnel confirm classifications, dispositions and overrides
Controlled Shipment Release
Release or hold with full approval audit history
| Agent | Responsibility |
|---|---|
| Master Quality AI SupervisorSupervisory | Coordinates specialised quality agents |
| Product Specification AgentProduct & Pre-Production | Analyses specifications and requirements |
| SOP Intelligence AgentProduct & Pre-Production | Assists with inspection procedure development |
| Pre-Production Validation AgentProduct & Pre-Production | Evaluates PPV readiness and findings |
| Golden Sample AgentProduct & Pre-Production | Supports reference-sample consistency |
| Inspection Planning AgentInspection | Assists inspection preparation |
| AQL Advisory AgentInspection | Explains approved sampling requirements |
| Computer Vision AgentInspection | Analyses product photographs |
| Defect Classification AgentInspection | Suggests defect categories and severity |
| Inspection CopilotInspection | Assists inspectors during inspection |
| Supplier Quality AgentSupplier & Corrective Action | Evaluates supplier quality trends |
| Factory Risk AgentSupplier & Corrective Action | Identifies manufacturing risk patterns |
| Root Cause Analysis AgentSupplier & Corrective Action | Investigates recurring defects |
| CAPA Intelligence AgentSupplier & Corrective Action | Supports corrective action evaluation |
| Predictive Quality AgentSupplier & Corrective Action | Identifies emerging quality risks |
| Compliance Intelligence AgentAssurance & Governance | Identifies regulatory and specification concerns |
| Anomaly Detection AgentAssurance & Governance | Detects unusual inspection patterns |
| Evidence Integrity AgentAssurance & Governance | Identifies incomplete or inconsistent evidence |
| Quality Reporting AgentAssurance & Governance | Prepares inspection summaries |
| AI Governance AgentAssurance & Governance | Monitors agent recommendations and permissions |
AI advises; deterministic software decides statistical acceptance; authorised people approve and release.
AI Advisory Layer
Proposed specialised agents analyse, suggest and explain — never decide
Deterministic Quality Engine
AQL calculations and mandatory quality rules in auditable software logic
Authorised QA Approval
Qualified personnel confirm classifications, dispositions and overrides
Controlled Shipment Release
Release or hold, with full approval audit history
It informs quality professionals. It never overrides them or the standards.
AI Computer Vision & Automated Defect Intelligence
Proposed AI capabilityAn AI vision architecture designed to help inspectors identify visible quality anomalies.
Computer vision findings can be correlated with approved specifications and inspection checkpoints. Inspectors retain responsibility for validating AI-generated observations and confirming actual defect classifications.
The objective is to improve consistency, reduce repetitive inspection work and support earlier identification of recurring manufacturing problems.
AI-Powered Supplier & Factory Quality Intelligence
Proposed AI capabilityInspection data consolidated across suppliers, factories, products and production periods.
The resulting dataset provides a foundation for advanced supplier-quality analytics. By examining historical quality performance, AI can help identify suppliers or product categories that may require increased inspection attention — moving businesses from reactive defect management towards risk-based quality planning.
Intelligent Reporting & Inspection Certificates
Structured reporting that reduces manual preparation of inspection documents.
AI-assisted drafting, human approvalAI-assisted report generation can consolidate inspection observations into structured draft narratives while preserving the original inspection evidence and deterministic quality results. Final approval remains with authorised personnel under established quality governance procedures.
Controlled Shipment Release
Inspection outcomes connected directly to shipment authorisation — so goods don’t move without the required approvals.
Enterprise Cloud Architecture
A modular cloud architecture on Railway that separates user interaction, business logic, background processing, identity and evidence storage.
Dedicated services for identity, business logic, background processing, data and evidence.
Users: QA Managers, Inspectors, Clients
Web portal and offline-capable Android inspection workflows
Keycloak Identity (OIDC + PKCE)
Secure authentication, roles and tenant isolation
Web Portal & API Service
Business logic, workflows and the deterministic AQL engine
Worker Service
Background processing, report generation and synchronisation
PostgreSQL & Cloudflare R2
Quality records, audit trails and durable evidence storage
| Component | Technology |
|---|---|
| Application hosting | Railway |
| User portal | Web application |
| Backend | Dedicated API service |
| Background processing | Worker service |
| Database | PostgreSQL |
| Identity management | Keycloak |
| Authentication | OpenID Connect |
| Login security | PKCE |
| Evidence storage | Cloudflare R2 |
| Data protection | Encrypted communications and controlled access |
| Integration architecture | Secure APIs |
Centralised Quality Management Dashboard
Quality information in one management environment. The views below are illustrative mock-ups of the eleven dashboard modules — not live data.
Specifications, SOPs and product readiness.
Readiness combines approved specification, SOP status, golden sample and PPV completion.
Illustrative mock-up. Products, lots, suppliers, factories and values are generic examples to show the interface — not records from a live deployment.
Commercial Applications
Manage product specifications, supplier inspections and shipment approvals across multiple countries.
Establish consistent quality processes throughout production.
Digitalise inspection execution, evidence collection and client reporting.
Maintain product-quality consistency across distributed manufacturing networks.
Improve visibility and control over pre-shipment inspection activities.
Manage recurring inspections, supplier performance and defect intelligence.
Provide structured inspection and quality governance services to multiple customers.
Why Loretto AI Labs?
PQIMP demonstrates Loretto AI Labs' ability to transform complex business requirements into secure, integrated enterprise applications.
Rather than forcing businesses to adapt their operations to rigid, generic software products, Loretto AI Labs designs technology around their actual processes, compliance obligations and commercial requirements.
Transform your manufacturing and sourcing operations with an intelligent quality management platform designed around your business. From pre-production validation to final shipment release, Loretto AI Labs can help your organisation establish a connected, transparent and AI-assisted quality assurance ecosystem.
Whether you manage a single manufacturing facility or an international network of suppliers, our enterprise software and AI engineering capabilities can be customised to your requirements. Build a smarter, more accountable quality management operation with Loretto AI Labs.