Skip to main content
UNSW Civil Engineer · ex-AECOM · 1+ year of agentic AI delivery
AI at Transdatos

An AI lab for industries software left behind.

Agentic AI in production every day — used as a multiplier on senior engineering judgement, not a substitute for it.

The gap § 01

What most AI work is missing.

Most AI work in Australian enterprise right now is one of three things:

  • A Copilot rollout the board can point to in a slide.
  • A chatbot bolted onto unstructured SharePoint data that hallucinates the moment a real question is asked.
  • A vibe-coded prototype that hits the wall around the 80% mark — the foundation too fragile to add features.

Each of those skips the part that makes AI compound: a real data foundation, an architecture that survives contact with production, and senior engineering judgement on what to build versus what to delete.

Transdatos closes that gap. We design and build AI-driven platforms that connect to the systems your industry already runs on, get stronger as more of the team plugs in, and keep working a year after launch. The result is not another confined product to log into. It is a platform that compounds.

How we work with AI § 02

Agent-led delivery.

Agents are the unit of execution

Transdatos is structured as an agent-led delivery firm — an orchestrated agent team running under production discipline, with a technical-cofounder agent driving architecture, the founder carrying judgement, and senior humans engaged on demand. The default unit of execution is the agent team — Claude Code, Codex, and the rest of the modern stack — operating against your problem. Every project demo includes short walk-throughs of the agent loop in action.

This is the structural answer to the $1M / 18-month consultancy quote. We do not throw cheap labour at the problem; we refuse to make labour the unit of work at all.

Some of what we deliver as projects today, we expect to operate as agent organisations we run for our clients tomorrow.

Frontier models, applied — not trained

Transdatos applies frontier models — Claude, GPT, Gemini — and builds the agentic systems, retrieval architectures, and data foundations around them. We do not train base models or fine-tune as a service.

MCP, RAG, and tool-use — where they earn their place

Model Context Protocol (MCP) tooling and retrieval-augmented generation (RAG) get used when they solve a real problem — not because they read well in a deck. We design the integration to the data and tools your team already uses, so the AI answers questions against the actual workflow rather than a snapshot of it.

A real data foundation underneath

Most AI failures are data failures wearing an AI costume. We fix the foundation first. Pipeline design, restructuring, agentic mining of unstructured data, migration to an architecture a real product can sit on — done before the AI layer goes anywhere near production.

Visible leverage, not hidden process

Bi-weekly demos showing real working software. Async-first between demos. No surprises at handover. The repository runs in your GitHub org at launch; the cloud deployment runs in your account. You own the IP. We earn from delivery and partnership — not from extracting it.

What you can buy § 03

Four offers under one cadence.

Each one has an AI angle; each one starts with the cadence — Hear → Lay → Shape → Ship → Compound.

TD · BUILD

Custom AI-Native Applications

Design and build of a custom web application, with AI and your data wired in from day one. Premium interface, real engineering depth, agentic delivery as the speed multiplier. The flagship.

Every Custom Application starts with a paid Discovery Sprint — one to two weeks, fixed fee. Outputs: solution architecture, delivery plan with milestones, fixed-fee build quote, risks and assumptions.

low-to-mid 6 figures AUD fixed-fee
TD · PARTNER

Product Partnership

The long-term engineering and AI team after the build. Continuous feature work, AI-layer operations, product maintenance, and — where it makes sense — co-build of internal IP into a market product.

This is where Transdatos becomes your outsourced AI lab — somewhere you send the hard problems, somewhere that builds the product, somewhere that operates it once it is live.

$30k+ AUD/month 3-month minimum
TD · DATA

Data Foundation

Agentic-AI-led restructuring of data that has outgrown spreadsheets and SharePoint. Pipeline design, query work, mining of unstructured data, migration to an architecture a real product can sit on.

This is not a Power BI service. The framing is agentic-AI-led data engineering. If a dashboard is part of the platform, fine. If a dashboard is the deliverable, this is not the right offer.

$30k–120k AUD fixed-scope
TD · ADVISE

AI Strategy, Advisory & Enablement

A seat at your strategy table, written strategy work, or hands-on workshops that teach what we ship — Claude Code masterclasses, data architecture for AI, executive AI-readiness sessions.

$6k–40k AUD per engagement
The cadence § 04

From problem to platform.

STAGE WHAT HAPPENS
§ 01 Hear the problem Intake, qualifying call, paid Discovery Sprint. The real problem usually sits underneath the surface ask.
§ 02 Lay the foundation Data, structure, architecture, integrations. Fixed first, before features.
§ 03 Shape the experience UI-first prototype using HTML and agentic AI. Iterated and signed off before any backend gets built.
§ 04 Ship the product Fixed-fee build, bi-weekly demos, the agentic delivery loop visible throughout.
§ 05 Compound the platform Product Partnership phase. The product becomes a platform; the platform connects to the rest of the industry's systems.

Five principles sit underneath

  1. § 01 Build the platform, not the product.
  2. § 02 Foundation before features.
  3. § 03 Visible AI leverage, not hidden process.
  4. § 04 Senior judgement on demand, not throughput.
  5. § 05 Partner, do not extract. Client owns the IP.
What we will not do § 05

Boundaries are the position.

  • Train or fine-tune base models.
  • Build “a Copilot for X” as a standalone deliverable.
  • Sell pure prompt engineering as a service.
  • Bolt a chatbot onto unstructured SharePoint data and call it AI.
  • Run AI delivery without fixing the data foundation first.
  • Make labour the unit of work. No pyramids of graduates billing hours, no hourly time-and-materials, no padded headcount on the proposal.
Primary CTA

Tell us the problem.

Press to speak, or type. The AI synthesises what you said into a structured problem statement before the call — so the first conversation is a working session, not a discovery interview.

Tell us the problem