Abstract: 

We are living in a time of big data, with multiple sectors gearing up for digital transformation and employing AI solutions that promise to effectively leverage this apparent data surplus. But in reality, many systems are challenged by having low observability - a real-world constraint that will remain constant due to the difficulty in tracking the billions of moving parts in today’s global networks. AI solutions capable of making sound decisions with sparse data will be the keys to cracking this data paradox.

At PROWLER.io, we utilise probabilistic modelling as an underlying foundation for decision making. As probabilistic models rely on principled mathematical methods that do not require large data sets to make accurate predictions, they are entirely suitable for operating in data sparse settings - and importantly, they quantify the uncertainty associated with each prediction. Integrating these models with other machine-learning techniques allow us to identify the most robust decision for a range of business scenarios, particularly in Supply Chain Management. We are helping businesses optimise decisions to align with their strategic priorities and make the right decisions at the right time.

In this talk, Sarah will be motivating the need for data-efficient solutions drawing on real world examples from our work in supply chain management. She will also be talking about the challenges we encountered - such as trying to forecast quantities where the input data is of a different type than the prediction - and what we learned when applying VUKU to complex, dynamic systems with little available data.

Bio: 

Sarah is both an engineer and a researcher, with experience in bioengineering and health economics. Growing up between the UK and Australia, she completed her PhD in Germany. As a data scientist, she bridges the research, software engineering and product teams to identify how best to build application software.