Custom AI & Machine Learning.
AI built around your data, not someone else's SaaS template.
The problem
You have heard that AI can transform your business, and you are probably right — but every vendor is selling a generic tool that does not fit your workflow. The ones that do fit are enterprise platforms priced for companies ten times your size.
What we build
Every AI engagement starts from your specific problem, not from a product we are trying to sell:
Predictive models — forecasting, anomaly detection, classification, trained on your historical data
Document intelligence — extracting structured data from contracts, invoices, emails, and PDFs
Agentic workflows — AI that can reason about multi-step tasks and take actions in your systems
LLM integrations — connecting language models to your data with retrieval augmentation, structured output, and guardrails
Automation — replacing repetitive human tasks with AI-powered classification, routing, and extraction
What we do not do
We do not resell other vendors' tools with a markup. We do not build chatbots that hallucinate. We do not take on projects where AI is not the right answer — sometimes a well-structured database query is worth more than a language model.
Pricing
AI engagements typically range from AUD 25,000–150,000 depending on scope and duration. Every project starts with a paid discovery week (AUD 5,000–10,000) so we both know what we are getting into before committing.
Related work
We have shipped
this before.
Manufacturing · Client work
Predictive Maintenance Pilot
Sensor analytics on factory line equipment
- duration
- 8 weeks
- lines after pilot
- 3
- downtime reduction
- −71%
Read the case study →
Legal services · Client work
Document Intelligence Pipeline
Structured data extraction from unstructured PDFs
- accuracy
- > 95%
- contracts/week
- ~40
- time per contract
- 6h → 25min
Read the case study →
E-commerce · Client work
Support Ticket Triage Classifier
Auto-routing inbound customer messages by intent + urgency
- mis-routes
- 12% → 3%
- CSAT change
- +8 points
- first response time
- 4h → 18min
Read the case study →
Internal · Our build
kt-agent
The conversational agent on this very website
- monthly hosting
- < US$10
- first-token latency
- < 500ms
- lines of TypeScript
- ~250
Read the case study →
Try it yourself
Live demos in your browser.
Industrial AI
Predictive maintenance
A live factory floor with 6 machines and a working anomaly detector.
Try it →
Document intelligence
Structured data extraction
Paste messy text. Get clean JSON out.
Try it →
Customer support automation
Sentiment & urgency triage
Type a customer message, see how we would route it.
Try it →
Forecasting & analytics
Quick forecast
Type 6 numbers, see the next 3 with a confidence band.
Try it →
FAQ
Common
questions.
Do we need a lot of data to start?
Depends on the problem. Some projects need thousands of labelled examples. Others — especially document extraction and classification — can work with 50 good examples and a well-designed prompt.
Which models do you use?
Whatever fits the job. We have shipped with Gemini, Claude, GPT-4, open-source models, and classical ML (XGBoost, scikit-learn). The model is a component, not the product.
What happens after handover?
Every AI engagement includes optional ongoing support. Models drift, data changes, and edge cases emerge. We monitor, retrain, and improve based on what we learn from production.
You might also need
Data
Data Engineering & Analytics
Pipelines you can trust — from messy spreadsheets to clean dashboards.
Learn more →
Software
Software Development
Internal apps that fit how your team actually works.
Learn more →
Consulting
Consulting & Strategy
Honest technical advice — even when the answer is "don't build that".
Learn more →
Ready to talk
Let's start with
your problem.
Every engagement starts with a conversation — not a sales pitch. Tell us what is not working and we will tell you what we would do about it.