Our Start-up of the Week is AMSimp, which is creating open-source software that leverages machine learning to improve numerical weather predictions.
Earlier this year, Dogpatch Labs hosted the Patch summer accelerator, which invited talented young people between the ages of 16 and 19 to receive mentorship from Irish entrepreneurs and tech veterans.
The accelerator – which aimed to pick out and prepare future entrepreneurs, scientists and engineers to help them create impactful companies, technology and research – played an important role in the foundation of AMSimp.
‘The problem we are solving is that weather prediction is not all that it could be’
– CONOR CASEY
Co-founded by Conor Casey and Laura O’Sullivan, the start-up sees an opportunity to improve the methods used to predict the weather. O’Sullivan, who is a self-taught machine learning developer, has previously won prizes for exoplanet detection and cervical cancer screening research. The teenager from Cork is currently studying computer science and mathematics at the University of Edinburgh.
Casey, who serves as CEO and head of forecasting at the start-up, is entering his sixth year at secondary school, where he has gained an interest in physics, meteorology, mathematics, chemistry and programming. He has previously conducted research on light pollution in Irish towns.
What’s the weather?
AMSimp is creating open-source software that leverages machine learning to improve numerical weather predictions.
“The problem we are solving is that weather prediction is not all that it could be,” Casey told Siliconrepublic.com. “For example, you may remember that Hurricane Sandy hit New York in 2012, even though it was not even predicted to reach the mainland.”
Musing on why weather forecasting can be so difficult, Casey said: “Forecasting today is done by physical modelling. This means we know the equations that govern weather systems, but we can’t solve them exactly. Instead, we approximate them.”
“Weather agencies take reams of raw data, feed it into supercomputers and just number crunch and number crunch. Weather forecasting has become increasingly accurate, but it takes so much computing power that we can’t tell with complete certainty if it’s going to rain.”
To try and tackle the problem, AMSimp takes weather records from the past decade and processes them through its models.
“We find weather patterns and forecast parameters such as temperature, humidity and wind speed,” Casey said. “These models are then used to make better predictions of the weather – following which, post-performing can be done in order to extract the most useful information from the forecast and to visualise the results.”
Early on in AMSimp’s journey, Casey said that the team found it challenging to talk to experts and researchers in the field of meteorology to get feedback and advice. But the start-up eventually overcame this hurdle and is now working with researchers from Met Éireann to get more detailed feedback.
Because it is creating an open-source project, AMSimp hopes to provide its technology to as many users as possible to improve weather forecasting tools for everyone.
“Our product is currently in early development,” Casey said. “Using our present model, we can provide a more accurate forecast, whilst using less computational power. Currently, we can provide 80pc of the accuracy of a 48-hour forecast.”
Existing 48-hour forecasts rely on physical models and supercomputers. Casey acknowledged that AMSimp has plenty more work to do, but said that he is “confident” that machine learning models could eventually become the cutting edge of weather prediction.
The start-up is not currently seeking funding, though Casey has not ruled it out entirely for the future.
“I am relatively new to the start-up scene in Ireland, and to start-ups in general,” he said. “There are fantastic opportunities and support available in Ireland to get your business off the ground. It is no longer necessary, in my opinion, to move to a place like San Francisco in order to be successful in the start-up world.”
Want stories like this and more direct to your inbox? Sign up for Tech Trends, Silicon Republic’s weekly digest of need-to-know tech news.