No code. No data. No servers. How Marino Software became a ML company.

Marino Software are one of the longest standing software consultancies in Ireland with almost 20 years under their belt they have seen their fair share of ups and downs. As part of our AIFightsBack webinar series CEO Keith Davey talks candidly about how they have morphed into a machine learning company helping their clients deliver exceptional customer experience. They have a pretty impressive list of clients who are testament to this service and it includes telcos, banks and retailers. Both slides and the webinar are available below.

A Journey to Machine Learning (ML)

Marino’s journey began in 2015 when their software developers started to upskill on cloud based solutions and building out their own solutions. At this time, systems were simplified and the range of free open source platforms such as TensorFlow were growing. This allowed the company to scale quickly as hardware was also expensive. It’s something that Keith reckons almost 6 years on has made the whole world of ML more accessible than ever to business.

In fact, now they are using serverless computing with artificial intelligence systems as a service, directly, without running dedicated servers in the cloud. There are many cost benefits of operating this way and Marino are now spinning out a sister company called Netzer to focus solely on cloud solutions for the telecom market. Another major benefit of morphing into a ML company are attractive R&D tax credits available. ML is considered a piece of experimental development allowing them to claim 25% of cost back. It’s a nice incentive explains Keith, especially since the work can and does often fail.

Journey to Machine Learning (ML) with Keith Davey, Marino Software (Slides)

A Journey to Machine Learning (ML) - a Software Perspective (AIFightsBack Series)

A Journey to Machine Learning (ML) with Keith Davey Marino Software (Video)

What does ML mean to Marino?

It’s mainly two things: Natural Language Processing (NLP) and Computer Vision driven. They have been working on voice recognition and voice synthesis projects and stuff around image detection and image classification. There are lots of others applications of ML out there but right now their focus is these two and Keith shared three very different use cases with us.

Permanent TSB – Banking

Marino created an identity verification app using open source platforms TensorFlow, Tesseracts and Amazon Rekognition. The app allowed the user to take a front facing photo and get instantly verified to within a 97% accuracy. This system has since advanced to operating completely in the cloud and has opened up the banking and telecom sectors to them.

iD Mobile – Mobile network

iD Mobile is a mobile virtual network operator operating in the United Kingdom. Marino developed a ML chatbot using entirely hosted ML tools. Built on the Google platform it offered customers a phone upgrade as a bot, largely reducing the number of calls to their call centre.

VERO – Social media network

Vero is a social media platform who were an overnight success gaining millions of users. As with any user driven platform they have challenges with adult content being uploaded and some of this included child pornography. Marino had to build a tool to automate the detection of illicit content.

They used prebuilt models from Yahoo in TensorFlow chained with Google age detection models to detect suspect content and this was then submitted to the FBI to deal with. The solution enabled them to reduce the amount of humans on their side involved in this detection which is really important to the team.

Voice Recognition Software

The final use case that Keith shared was a voice recognition software for Roisin Foley, a mum who has Motor Neuron Disease. This project was part of the Big Life Fix tv show recently aired on RTE. Marino were engaged to develop software to help Roisin communicate with her kids with her own voice after she will lose it due to the illness. First, they recorded her voice and built a custom app, they combined this with some IoT in her home that her kids could easily use.

It was this combination of her own voice recordings and audio mimicry technology Lyrebird that allowed them to build a far superior voice synthesis tool than anything else on the market. The results where astounding for Roisin and her family and this project has kick-started Marino’s venture into conversational or voice design apps.

What’s the future of Marino?

The future for Marino is about focusing on delivering value for customers. This means chaining systems together to deliver intelligence solutions for customers.

Further webinars

Review the entire series on youtube.