01 / AI & ML
AI & Machine Learning Systems
RAG pipelines, vector search with CLIP embeddings, document understanding, computer vision, LLM orchestration. Production deployments, not notebooks.
PythonOpenAIGeminipgvectorCLIP
AI-accelerated software delivery · Cape Town → the Gulf
Software systems that ship, not slide decks.
One senior engineer, AI-accelerated: the output of a small team, shipped by the person you brief. W-Enterprises designs and delivers production AI/ML and full-stack systems from specification to production in around three weeks, including bilingual Arabic and English platforms for the Gulf.
01 / Practice
What we actually build.
02 / Engineering
Full-Stack Engineering
Type-safe applications on Next.js, React, Vue, Node, .NET and Go. Serverless-first deployments on Vercel, Cloudflare, Firebase. Speed without sacrificing correctness.
Next.jsReactTypeScriptGo.NET
03 / SaaS
SaaS Platform Build-Out
Multi-tenant architectures, subscription billing, identity, admin tooling, regional payment integrations. End-to-end from specification to production within compressed delivery windows.
FirebaseFirestoreMoyasarStripePostgreSQLMulti-tenant
04 / Discovery
Technical Discovery & Documentation
Solution architecture, PRDs, technical specifications, requirements traceability. The formal paperwork that lets serious engagements move forward with confidence on both sides.
ArchitecturePRDsSpecificationsMermaid
02 / Built for the Gulf
Built for the Gulf, not just translated for it.
Bilingual Arabic and English platforms with correct right-to-left interfaces, shipped for the Saudi and Gulf market. PDPL data protection, ZATCA e-invoicing and Moyasar payments are handled as first-class requirements, not afterthoughts.
هندسة برمجيات وذكاء اصطناعي للسوق الخليجي
Bilingual
Arabic / English + RTL
Native AR/EN interfaces with correct right-to-left layout, bidirectional text and Arabic typography. Not a translate widget.
Compliance
PDPL-ready
Data-protection-aware architectures: sensitive fields handled to Saudi PDPL expectations by design.
E-invoicing
ZATCA-aware
Invoicing and document flows built with Saudi ZATCA e-invoicing requirements in mind.
Payments
Moyasar + regional
Local payment integration with Moyasar, alongside Stripe and PayPal for cross-border billing.
- / Principles
How the practice operates.
i.
One senior engineer.
No account managers. No project handoffs. The person you brief is the person who writes the code, designs the architecture, and ships the system.
ii.
Production-first.
Every deliverable lands in production with monitoring, observability, security review and CI in place, not a demo on a developer's laptop.
iii.
Documented.
Every engagement comes with formal proposals, specifications, and traceable requirements. The client always knows what they're paying for and when it will be delivered.
iv.
Honest scope.
Out-of-scope items are named explicitly upfront. Change requests are quoted separately. No surprise invoices, no scope creep on the engineering side.
03 / Selected Work
Recent work.
2026
AI Invoice Parsing PlatformKSA · FinOps
Multi-tenant SaaS for automated invoice ingestion and field extraction, with hybrid OCR plus LLM normalisation and bilingual AR/EN.
PythonGeminiGoogle OCRPostgreSQL
Context
Finance-operations platform for the Saudi market, bilingual Arabic and English, with WhatsApp document ingestion.
Problem
Finance teams re-key invoices by hand across formats and two languages.
Approach
Google Document AI plus Gemini for hybrid OCR and field normalisation. Multi-tenant and stateless: documents stream through, nothing persists to disk.
Result
93.5%field-extraction accuracy on MVP scope
2026
BC Auto Workshop SiteZA · Automotive
Migrated a locked-in GoDaddy site-builder presence to a self-hosted, hand-built WordPress site with structured-data SEO and a performance pass.
WordPressPHPCustom ThemeSEO
Context
A Cape Town auto workshop stuck on GoDaddy's Website Builder, which cannot export its own site.
Problem
The builder locked in the content, throttled performance and gave little control over on-page SEO.
Approach
Captured the legacy site, rebuilt it on self-hosted WordPress with a hand-built theme (no page-builder bloat), AutoRepair and FAQ structured data, 301-preserved URLs, and a local-SEO and speed pass.
Result
Off GoDaddyself-hosted WordPress, custom theme, technical SEO. View the site →
2025
Trademark Similarity EngineKSA · IP Rights
Multi-modal vector-search system that screens trademarks for visual and textual similarity, with a scraping pipeline and colour-distribution matching.
PythonCLIPpgvectorFastAPI
Context
Trademark clearance for registration in the Saudi IP market.
Problem
Manual clearance is slow and misses near-identical marks across visual and textual variation.
Approach
CLIP ViT embeddings for visual similarity, OCR plus a vision-language model for text, histogram and CIELAB colour, and fuzzy name matching, served from FastAPI over PostgreSQL with pgvector.
Result
94%top-5 visual-similarity accuracy; 87% combined visual and textual
2025
IoT Camera MiddlewareUS · Surveillance
Reverse-engineered the proprietary UDP protocol of low-cost IoT cameras to expose standard RTSP and MJPEG streams for off-the-shelf NVR systems.
GoProtocol RERTSPFFmpeg
Context
An integration layer letting cheap V720 and Yi-class cameras join standard surveillance stacks.
Problem
The cameras only speak a closed protocol through a poor first-party app, so they cannot join standard NVR systems.
Approach
Reverse-engineered the UDP protocol and built Go middleware that re-exposes the feeds as RTSP and MJPEG via FFmpeg.
Result
Open streamsstandard RTSP and MJPEG from a closed protocol
2024
Storefront PlatformDE · E-Commerce
Custom Shopify-alternative storefront with a visual page builder, featured collections and a merchant onboarding flow.
Vue.jsGoPostgreSQL
Context
A bespoke commerce platform for merchants who needed more control than hosted storefronts allow.
Problem
Off-the-shelf storefronts constrained layout and merchant onboarding.
Approach
A Vue.js front end with a visual page builder over Go services and PostgreSQL.
Result
Visual buildermerchant-configurable storefronts
Building something similar? Start a project →
- / The Practice
Led by an engineer who writes the code.
W-Enterprises is led by Wayne Fourie, a senior engineer with over a decade of production experience across AI/ML systems, full-stack web applications, and embedded middleware.
Recent engagements span AI-powered document parsing platforms, trademark similarity engines using CLIP embeddings and pgvector, reverse-engineered IoT camera middleware, and multi-tenant SaaS deployments for the Saudi Arabian HR-tech market.
The practice is small by design. The trade is between scale and rigour, and we choose rigour every time.
Wayne Fourie
Founder and Principal Engineer
- / How it works
How an engagement runs.
01 →
Discovery
We pin down the problem, the constraints and what success looks like. You get a senior read on scope and risk before anything is committed.
02 →
Spec and proposal
A written specification and PRD with traceable requirements. Scope is fixed and out-of-scope items are named, so both sides know exactly what ships and when.
03 →
Build
AI-accelerated delivery on serverless architectures, with monitoring, security review and CI in place from the start. Regular check-ins, no black box.
04
Ship and handover
The system lands in production with documentation and a clean handover. Code ownership and licensing are agreed in the proposal.
04 / Technology
The full toolkit.
Production-grade work calls for production-grade tools. The stack below reflects what we've shipped, not what's been touched in a tutorial. The specialised groups come first; the rest is one click away.
AI & Machine Learning
PyTorchTensorFlowLangChainLlamaIndexCLIPHugging FaceOpenAI APIClaude APIGoogle GeminiWhisperRAGVector SearchpgvectorChromaDBPydanticDocument AI
Specialized Domains
Multi-tenant SaaSOCR & Document AIComputer VisionVector SimilarityProtocol RERTSP / MJPEGFFmpegBilingual (AR/EN)RTL RenderingPuppeteer / PlaywrightWeb Scraping
Regional & Payments
Moyasar (KSA)StripePayPalPDPL ComplianceZATCAArabic LocalisationSARS / VAT
+ Show the full stack: languages, frontend, backend, data, cloud
Languages
PythonTypeScriptJavaScriptC#GoJavaPHPSQLHTMLCSSBashPowerShell
Frontend
Next.js 14ReactVue.jsNuxtReact NativeExpoElectronTailwind CSSshadcn/uiRadixZustandTanStack QueryRecharts
Backend & APIs
Node.jsExpressNestJS.NET / .NET CoreASP.NETFastAPIFlaskGoRESTGraphQLgRPCWebSockets
Data & Storage
PostgreSQLpgvectorSQL ServerMySQLMongoDBSQLiteRedisFirestoreSupabaseElasticsearchPineconeWeaviate
Cloud & Infrastructure
Google CloudFirebaseAWSAzureVercelCloudflareNetlifyDockerKubernetesTerraformGitHub ActionsGitLab CILinux
- / FAQ
Common questions.
What does W-Enterprises do?
W-Enterprises is a solo software consultancy that builds production AI/ML and full-stack systems end to end: RAG and LLM applications, computer vision, OCR and document AI, and multi-tenant SaaS. One senior engineer handles architecture through deployment.
Who is Wayne Fourie?
Wayne Fourie is the founder and principal engineer of W-Enterprises, a senior engineer with over a decade of production experience across AI/ML, full-stack web, and embedded systems, based in Cape Town, South Africa.
How can one engineer deliver platform-scale work?
By leveraging AI heavily across the build and working serverless-first, the practice delivers the output of a small team from one senior engineer, without agency overhead or handoffs. A typical platform goes from specification to production in around three weeks.
Do you build bilingual Arabic and English software and handle Gulf compliance?
Yes. Shipped work includes bilingual Arabic and English platforms with right-to-left interfaces for the Saudi and Gulf market, accounting for PDPL data protection, ZATCA e-invoicing, and Moyasar payment integration.
Which markets and timezones do you work in?
Remote-first from Cape Town at UTC+2, which overlaps the Gulf and Europe through the working day. Active engagements span South Africa, Saudi Arabia and the Gulf, Germany, and the United States.
How do engagements work?
Work begins with technical discovery and a written specification. Scope is fixed and documented, and out-of-scope items are named upfront. Code ownership and licensing are agreed in the proposal.
05 / Contact
Have a system that needs building?
The practice takes on a limited number of engagements concurrently. If your project is a strong fit, you'll have a response within one working day. Briefly: what you're building, when it needs to ship, and what success looks like.
Email
WhatsApp
Based
Cape Town
South Africa · UTC+2
South Africa · UTC+2
Operating in
ZA · KSA · DE · US
Remote-first
Remote-first
Availability
By engagement
Reply within 1 working day
Reply within 1 working day