Water pipe breaks cause very severe damage. Big ones could cost millions of dollars to fix. 

If you just look at Sydney alone, we have over 20,000 kilometres of water pipe. We want to know if we can fix those infrastructures before the break happens.

I was trained as a rocket scientist, and then at some stage I'm quite interested in AI. Data is everywhere and everyone is asking about how to make better use of data in day-to-day work or life. These are probably 60, 70 years old pipes. 

Since those pipes are getting older, we have more problems. It's the same as us.

The pipes are buried underground. Some could be a hundred metres down. The condition of the pipe is one of the mysterious parts. We don't know how bad the condition is. What is the thickness of the pipe that defines when this is going to burst? 

How to get that data, put it on a robot.

This particular sensor needs to touch the wall. In order to make good quality data, measure those and feed into your amazing algorithm.

The prediction power of AI and data science is remarkable. You can do the what if scenario planning. So, what if this happens to identify which pipes need attention?

All AI data science is probability based eventually looking into causal effect of a certain data stream to an outcome. 

So you see the status of the red one is leak. We have been working with Sydney Water and in 18 month trial we saved more than 10,000 megalitre water. That's more than 4,000 Olympic pool size.

And we saved $20, $30 million. And when you drink water in the morning, you still have water. So we are basically building a digital version of this infrastructure. Data science is a practical science. We create the digital twin for the entire city. Essentially it replicates the physical world.

Then you can play in that and see how the effects look like before you apply to the real world. We want to save water. Water is a precious resource.

We are looking into the mega cities like New York, London, Tokyo, to put the technology in much wider use. Data is everywhere. If you just mine those data, there's potential to do amazing things.