One feature, one owner, end to end. You will not be handed a ticket cut out of someone else's design — you take the thing from spec to shipped, and "done" means it shipped and produced a result.

You get your own agent, and you are accountable for what it does. Every agent on our platform is registered against a named human orchestrator; without one it cannot be created, and every action it takes is logged. The agent widens what one person can build. It does not move where responsibility sits.

You will take a turn as Integration Owner. It rotates each release train, it is capped at one train, and while you hold it you do not claim features of your own — you hold the source of truth and make the ship call.

Independent QA can block you, and is not allowed to review its own work. A delay caused by a legitimate quality block is not counted against the person who was blocked.

If that sounds like more ownership than you want, this is not the right place. If it sounds like the job you have been trying to get for years, talk to us.

Company Description Tecco is a product and technology company building Homira, a comprehensive smart home platform spanning resident apps, cloud services, design tools, and in-wall devices. By owning every layer of the stack, Tecco enables deterministic, secure home automation and solves long-standing challenges in smart home design, configuration, and deployment. The company transforms home design into software, allowing rapid, scalable configuration that licensed electricians can install without custom programming on site.





Tecco also designs and manufactures its own KNX-certified hardware and operates as an authorised KNX Association training centre, with a strong focus on energy visibility and dispatch capabilities. The team is based in St Leonards, Sydney and is expanding across home intelligence, fleet, edge, energy, and design automation.

Role Description We are building the layer that turns "I'm going to watch a movie" into a room that is already right — behaviour recognition, spatial context, then a target state expressed as brightness, glare, temperature, privacy and safety bounds, before any device is told to do anything.

The design commitment we have made is that the language model understands; it does not actuate. Interpretation is the model's job. What actually moves in someone's house is decided by deterministic composition against device capability. You would own that boundary.

This line is at MVP. The behaviour engine is designed and partially built. It is not shipped, and we will not describe it as shipped.

First 90 days: take the scene synthesiser from design into something running against real homes; close the learning loop so that a resident nudging the lights down one notch actually updates their preference.

What we need: you can build systems that make decisions under uncertainty and still explain themselves. Applied ML, recommender or personalisation background is relevant; so is control systems. You care that a home is a white box the resident can see into — trust in a house is often one-shot, and one wrong action loses it.

This role is not: a research position. Nothing here gets to stay a paper.

📌 Home Intelligence Engineer (Sydney)
🏢 Tecco
📍 Sydney