Aurora-AI has developed a number of solutions for the air industry which are helping to improve operations, border security and passenger management throughout the airport environment. Our AI based facial recognition is powering the largest biometric roll out ever seen in the aviation segment and we are now applying our AI expertise to predictive analytics and analysis of scanned images for threat detection and classification for both passengers and baggage.

Air transport, with its ever-increasing passenger volumes and the huge numbers of competing factors, presents many forecasting and logistics challenges; AI is extremely well placed to tackle these, finding correlations in data without human direction. Whether it’s presenting live updates to inform decision making or forming part of the airport’s “digital twin” to generate insights into future performance, AI is becoming an essential part of airport planning and operations for our customers.

Working with Aurora-AI

Our consultants work closely with each customer to develop the solution and deliver it as an end to end service, which is not simply based around “off the shelf” technologies such as Azure ML or Watson. Our Data Scientists are on hand to deal with the critical step of data preparation “in house”. This means our clients do not have to be Deep Learning experts, with the finished article being delivered as an API for integration, a cloud-hosted service or a complete application with full interface, comms and storage.

Applied AIs for specialist solutions

Artificial Intelligence (AI) is a computer programme, typically incorporating an artificial neural network, that can learn to perform complex intellectual tasks.

At Aurora-AI we create Applied AIs – artificial intelligence to solve one specific problem, really well. Once created, it cannot be applied to another task. An example might be an AI that recognises faces. It won’t recognise cars, create recipes or predict the weather. But the next one we create could do any one of those things.

Creating, testing and delivering AI Solutions

Aurora-AI has built a high-performance cluster of GPU (Graphics Processing Units) nodes, which have been engineered specifically for Deep Learning. This platform provides numerous tried and tested ‘template’ neural network architectures, which all differ in size, types of neuron, layers and structure, because our experience shows that the neural profile of the best solution is not always immediately obvious.

This approach means we can test across many different neural network structures to quickly find the most efficient solution for any given task, differing from the more orthodox method of picking a structure based on previous experience of a similar problem and trying to make it fit.

The final packaged solution will run on standard hardware.

The five steps of AI Development

An initial discussion to assess aspects including the application, the current solutions and the type and volume of data available.

Our experts will review the requirement and analyse a data sample, producing a report that includes a costed proposal for the remainder of the programme.

The training data will be cleansed (where necessary), split into a number of sets for training and formatted ready for Deep Learning.

Our proprietary Deep Learning platform will be used to develop and optimise the AI.

The finished AI is packaged with the Aurora-AI Software Development Kit and provided ready for integration.