Lumin8te

Project Portfolio

Lumin8te's mission is to empower businesses and individuals with intelligent technology that elevates how they work and create.

● Case Study

TCSG Manufacturing Intelligence System

Knowledge platform and RAG solution for a TiO₂ (titanium-dioxide) pigment manufacturer (Perth, Australia).

The Challenge

Manufacturing TiO₂ is a complicated endeavour. The pain points of the client revolved largely around accessing data and specific information, knowledge attrition, operational bottlenecks, and onboarding and upskilling individuals.

The Solution

A RAG-based AI system built on the client's own data, with secure, tiered access. Operators get accurate, sourced answers in seconds instead of delayed information retrieval. Knowledge is centralised, consistent, and improves over time through a corrections workflow — with a path to live plant data for real-time insight.

What was built

  • Process-mapped knowledge AI agent — ask a question across any part of the process, get a sourced answer with references, configured to the operator's access level.
  • Corrections workflow & admin panel — user management, flagged answers, and an expert-curated corrections functionality so accuracy compounds over time.
  • Tiered, role-based access — the right knowledge to the right people, securely.
  • Cloud storage — production-grade database infrastructure.

Architecture & Stack

AI Engine & ArchitectureClaude Code · custom skills library · MCP servers (n8n, GitHub)
Orchestrationn8n · Mistral (OCR) · data ingestion · agent orchestration · retrieval pipeline
Data & StorageSupabase / PostgreSQL · pgvector vector store & embeddings
Development & InfrastructurePython · TypeScript · JavaScript · React · Next.js · APIs · Webhooks · Vercel · GitHub

Impact

  • Sourced answers in seconds, replacing long manual searches through scattered documents.
  • Plant knowledge centralised and consistent across shifts and experience levels.
  • Accuracy compounds over time through the expert-curated corrections workflow.
  • A foundation ready to extend to live plant (DCS) data for real-time intelligence.
“Working with Jared to produce an AI for the TiO₂ manufacturing industry has been an awesome experience. He was able to take our database of intellectual property in different formats and, in short time, configure the AI to extract targeted information in a highly efficient way. This is the way forward for any organisation which has a large amount of technical, financial and commercial information in digital form and needs to access it efficiently.”
Jon Howlett — Senior Consultant, Titania CS Global
“Jared set up our AI Manufacturing Intelligence System and did a fantastic job. Highly recommended.”
Dereck Becker — Director, Titania CS Global
● Case Study

Events Planning Portal

A bespoke events-planning solution for a luxury boutique events venue (NSW, Australia).

The Challenge

Every new booking triggered a 31-page Google Doc that users filled in by hand — rigid, impersonal, and jarring against a venue selling $100k+ weddings with 1–2 year lead times. Worse, staff then manually re-keyed every answer into their project management tool — double data entry, hours spent per event. Another challenge was missed ROI opportunities as the manual workflow made for a clunky upsell engine.

The Solution

A custom portal that replaces the document with a tailored journey — and quietly does far more than a document ever could. Conditional logic adapts each step to prior answers (whole sections skip when irrelevant), a contextual upsell engine surfaces the right add-on at the right moment, data stored in production-grade database infrastructure, and an educational layer guides first-time users. Auto-save means progress is never lost — even offline — and the output is structured, CRM-ready data with no re-keying.

What was built

  • 18-step conditional portal — replaces a 31-page document with a staged, adaptive journey.
  • Rules-driven upsell engine — add-ons trigger on guest count, catering, and event choices, systematising contextual selling earlier and data-driven.
  • Admin dashboard + content management — staff onboard couples and edit portal content and upsells without the need for a developer.
  • Automated transactional email — welcome, reminders, and submission alerts from a verified sending domain.
  • Data CSV export — one-click handoff into the venue's CRM (Tripleseat).
  • Signed single-use links, row-level-secured database, Australian Privacy Act 1988 compliance — consent, collection notice, and privacy policy built in.
  • Separate DEV/PROD environments, semantic versioning & one-click rollback — production-grade release discipline.

Architecture & Stack

AI Engine & ArchitectureRules-based upsell trigger engine
OrchestrationNext.js server actions + API routes
Data & StorageSupabase (Postgres) · Row-Level Security · isolated DEV / PROD projects
Development & InfrastructureNext.js 14 (App Router) · React · Tailwind · Framer Motion · Zustand · React Hook Form + Zod · Resend (email) · CSV · Vercel

Impact

  • Eliminated manual double data entry — structured output flows straight into the CRM.
  • Turned a 31-page PDF into a premium first touchpoint across a 1–2 year client journey.
  • Upsells systematised, not ad-hoc — the contextual selling behind a 15–50% revenue uplift, now earlier and data-driven.
  • Staff self-serve content and upsell edits — no developer in the loop.
● Case Study

Orbis

A multi-platform AI social content system for Lumin8te (in-house build).

The Challenge

Lumin8te needed a robust social media posting automation developed, to assist the organisation with its marketing strategy. Manual content creation can be a time-consuming endeavour, and so to free up that time to be able to work on more pertinent deliverables, an automated solution was needed.

The Solution

An end-to-end system that turns one topic row into multi platform-tailored posts, keeps the stakeholders in the approval loop, and publishes on demand to LinkedIn, Facebook, and Instagram from a single dashboard. Content is generated per platform and a React approval dashboard sits between generation and publish so nothing goes live unedited. On approval, publishing writes back to the log automatically.

What was built

  • Topic-driven content queue — one Google Sheet row delivers three platform-tailored drafts in a single generation run.
  • Multi-format post generation — text, single-image, quote-card, and PDF-carousel formats, each with format-specific prompt discipline and platform-native voice.
  • Native LinkedIn PDF carousels — server-side slide rendering (1080×1080 PNG per slide → multi-page PDF) uploaded via LinkedIn's Document API. Swipeable in-feed, not a link-out.
  • Instagram native carousel pipeline — the same slide assets re-used as an Instagram swipeable carousel via the Meta Graph API's create → aggregate → publish flow.
  • Facebook quote-card publishing — brand-consistent single-image posts via the Meta Graph API.
  • Human-in-the-loop approval dashboard — a Next.js UI for reviewing, editing, and approving each draft before publish. Zero-click auto-posting does not exist, by design.
  • Approval-triggered publishing — approval fires the platform Publisher workflow, which updates the sheet and dashboard status atomically on success.
Text Image Quote Card PDF Carousel IG Carousel

Architecture & Stack

AI Engine & ArchitectureAnthropic Claude — per-platform system prompts encoding voice, format discipline, and hashtag conventions
Orchestrationn8n — Content Generation workflow + three platform Publisher workflows (LinkedIn, Facebook, Instagram)
Data & StoragePostgreSQL (Vercel Postgres) · Google Sheets
Development & InfrastructureNext.js 16 · React · Tailwind · Prisma · Satori · pdf-lib (multi-page PDF) · PNG rasterisation · LinkedIn API v2 · Meta Graph API v25.0 (Facebook + Instagram) · Google Sheets API · Vercel

Impact

  • One topic → three platforms — a single topic row becomes three platform-tailored drafts in one run, eliminating triple-authoring.
  • Human quality gate — every post is reviewed and editable before it goes live; zero unreviewed posts, by design.
  • Native swipeable formats — LinkedIn and Instagram carousels render as native feed content, not link-outs — matching the formats that earn distribution.
  • Source-of-truth logging — posted text and date are written back to the sheet automatically on publish, so history is queryable in one place.

Selected work

More projects

AI Adoption

AI Adoption & Learning Portal

A gamified internal training platform built specifically to assist users in becoming comfortable with using AI tools. Curated learning paths and knowledge library, a peer based wins log so users can learn from each other, and a real-prize leaderboard to spark some healthy competition.

Next.jsSupabaseGamificationMicrosoft CopilotChange Management
Data & BI

Operational Reporting & Analytics

Reporting and analytics systems that improved visibility into business performance and gave leadership sharper, faster inputs for decision-making.

SQLPower BIQlik SenseSSRSData ModellingMicrosoft 365
AI Automation

Lead Generation & Outreach

An AI-driven lead generation and outreach system using personalised messaging and workflow orchestration to source and engage prospects at scale.

n8nAI PersonalisationApolloInstantlyResend

Why Lumin8te?

Building AI is the easy part now.

Anyone with the right tools can generate working code nowadays — that's no longer the hard part. But it still takes a lot of tangible and intangible factors to implement successful projects. That's where over fifteen years of experience of being in the trenches shows up.

01

I have been shipping real projects for over 15 years — so I know a lot about what derails them.

Change management, user adoption, scope creep, shifting expectations, ROI trade-offs. I build the thing that moves the business and get people to actually use it.

02

A user should be enjoying their experience.

Functionality and fit for purpose builds is obviously very important, but I also place a lot of emphasis on user experience (UX) and how the user interface (UI) can complement this experience.

03

Mapping business processes and data management.

With a lot of experience in business analyst functions, I place a lot of emphasis on the business process and on data. If these are inadequate at the start, a software or AI implementation is already on shaky ground to succeed.

04

Engineering rigour & responsible delivery.

Maintainable, modular systems built to run in production — with governance, security and compliance at the forefront of design thinking.

The tools are commodities. The judgement isn't.

Capabilities

The stack

Production-grade tools, chosen for reliability and maintainability.

Orchestration
n8n · Claude Code
AI Models
Claude (Opus / Sonnet / Fable) · GPT · Mistral OCR
AI Architecture
MCP · RAG pipelines · Claude Skills
Data & Storage
PostgreSQL · Microsoft SQL Server · Supabase · pgvector · Pinecone
Development
Python · Node.js · TypeScript · JavaScript · React / Next.js · APIs · Webhooks · GitHub
Growth & Outreach
Apollo · Instantly · Resend
Platforms & Infrastructure
Microsoft 365 · Google Workspace · Microsoft Copilot · Vercel

Meet the Founder

Jared is the founder of Lumin8te and a passionate advocate for practical, people-first AI implementations. With a background spanning over a decade in enterprise technology, analytics, and data-driven automation, Jared founded Lumin8te to bridge the gap between cutting-edge AI capabilities and real-world business outcomes.

His philosophy is simple: technology should serve people, and it should be an enjoyable and fulfilling experience for end users. That belief drives everything Lumin8te builds — custom solutions tailored to the unique needs of each client.

"I am passionate about creating products and experiences users don't forget—solutions that exceed expectations and create meaningful value for my clients' operations."

Jared Davey, Founder of Lumin8te

Let's build

Interested in finding out more or booking a free discovery call?