Ocado data chief: 'Machine learning skills are very hard to find - we have to compete with Google DeepMind'

Ocado's technology division - which employs over 1,000 developers, engineers, researchers and scientists - is having to compete with tech giants such as Google for machine learning expertise as it creates new artificial intelligence (AI) systems in-house.

Demand for data scientists is increasing among UK enterprises, and staff with PhDs in areas such as natural language processing, computer vision and deep learning techniques are highly sought after by certain companies.

"People with those skills are very hard to find, they are a finite set," said Ocado Technology's head of data, Daniel Nelson. "Google recruiting them all is one of the main [challenges] and DeepMind is a vacuum for very, very clever people."

"Yahoo Labs, DeepMind and Facebook are all trying to do very clever stuff to understand interactions with their consumers and we are very involved in that space," he explained.

Read next: How to get a job as a data scientist: What qualifications and skills you need and what employers expect

Ocado Technology is a customer of Google's and has used its APIs and open TensorFlow library to apply artificial intelligence to its contact centre processes.

Nelson said that, while machine learning experts may be in short supply, access to pre-built machine learning tools is 'democratising' use of the technology. Read next: Ocado to replace barcode scanning with AI 'vision' to speed packing processes

"It is a hot market," he said. "If you need somebody who has a PhD inside of natural language, or vision or these kind of deep learning methods, it is very, very hard. It is a very competitive market." Nelson added that the company is now looking overseas for data scientists in some cases.

"But what I would say that things like TensorFlow are democratising that a lot," he explained. "There is a lot of the learning and the research is already in there. The theorems and the mathematics is there, it is about you picking the model that works best.

"We do have a variety of PhD alumni working in our data science teams. None of them have got natural language as a background academically, so [access to machine learning tools] democratises that."

He added that the ability to access machine learning APIs in the cloud simplifies the creation of AI systems: "With NLP API you don't even need to be a scientist, you throw some text at it, you can do it on Google's website and it will tell you what this piece of text means, what language it is in, and how its syntax has been put together. That is just commodity machine learning."

"Tensor is just democratising it and removing some of the high end academics from it, so you just need to know how you train a model and help pick what the best model is," he said.

Copyright © 2016 IDG Communications, Inc.

Download: EMM vendor comparison chart 2019
Shop Tech Products at Amazon