Part 1 - Engineers Make The Best Data Scientists!

So don’t miss out on an opportunity to accelerate your career and revolutionize the way you think about your network and customer service quality.


Data analytics, machine learning and artificial intelligence are already the buzzwords of a revolution. Everything is AI now - every large company has a data analytics strategy, every start-up has ‘AI built-in’ if it seeks funding. The AI revolution has visibility and priority at Board and CEO level, so funding and resources for innovation is plentiful. We should of course be naturally sceptical of such hype, but in the world of telco CPE we believe there are good reasons to think the promise of AI is even bigger than currently perceived.

Harvard Business Review, the world’s most prestigious management journal, labelled ‘Data Scientist’ as “The sexiest job of the 21st century”. McKinsey, the global leader in management consulting, agrees and has invested massively from an early stage in building its own analytics capabilities. Its lead has now been followed by all the major international engineering consultancies.

For engineers with vision this creates incredible career opportunities, not just high salaries that data scientists command, but more interesting work that easily attracts the attention of senior management, giving access to projects and influence beyond your years of experience and seniority.

Yet telecom engineers, particularly CPE engineers, have not been prominent in the AI revolution. If you’ve missed out to date, at Axiros we believe CPE engineering is about to be changed forever by data and your time is about to come. And as a CPE engineer, you have 2 enormous advantages:

  • You already know the technical foundations of machine learning (even if don’t realise it!); and

  • Most importantly, you really know your stuff when it comes to CPE systems.

Because there’s simply no substitute for understanding what’s really going on when you’re trying to interpret complex data and help AI make sense of it.

 
The foundations of machine learning lie in Information Theory

AI has a reputation as a ‘black box’, and has more than its fair share of obscure technical jargon: support vector machines, connectionist models, bagging and boosting, ensemble and kernel methods etc. But in reality its time as a ‘black box’ has already passed, and it’s rapidly going to become just a standard part of the toolset every engineer should learn and use.

Fortunately for telecoms engineers, you have a huge, hidden head-start when it comes to making sense of machine learning – it’s based on Information Theory so you’re already familiar with its fundamental principles!

For example a Decision Tree, one of the most popular ML algorithms, operates simply by repeated discrimination based on estimates of entropy and information gain ratios. Concepts like Huffman Encoding, the Akaike Information Criterion, cross-entropy and the Kullback-Leibler Divergence all play central roles in the way various machine learning approaches operate.

Even sophisticated modern developments such as the ‘Deep Learning’ approach to Neural Networks, are, at their core, just large-scale applications of the ‘gradient descent’ technique of numerical approximation that is familiar to most engineering students.

So with the technical background to quickly learn the ‘black box’ techniques of machine learning, you’re ideally placed to make the most of your real advantage – your domain expertise.


Domain understanding trumps data

The rule of thumb that more data beats a better algorithm is widely acknowledged in data science. But what’s less well understood is that when it comes to delivering real, practical AI projects, domain knowledge is just as critical.

In the early days of data analytics much of the focus was on marketing applications, and ‘pure’ data scientists could work effectively where customer problems and data fields are relatively intuitive and understandable for non-specialists.

But as the focus turns increasingly to operations and engineering applications, there’s no substitute for real domain expertise. The hardest parts of a problem become specifying it properly and interpreting it reliably. How could a non-specialist even begin to comprehend the significance and subtleties of patterns in complex sets of moment-by-moment CPE data? As an engineer, you hold the key to understanding the defining steps: how to prioritise, frame and interpret analyses.

In fact McKinsey has recently identified the role of the ‘Analytics Translator’ – the person who understands both the technical / business domain and the capabilities of machine learning – as the critical role where organisations need to deploy their best talent. This is where CPE engineers can shine – not as the algorithm geek nor the person responsible for the nitty-gritty of “plumbing” data systems – but as the real driver of which problems should be tackled and how.

By knowing enough about machine learning to understand what’s possible, and working in collaboration with those who can handle the IT and data technicalities, you can command the application of all the AI horsepower to the technical challenges that only you understand properly.

In practice, machine learning is very much a practical ‘art’ not a theoretical ‘science’. An engineer’s pragmatic, problem solving mindset is therefore perfect for the translator and implementor role, where what matters is figuring out how to get things done.

Simply put, telco engineers are ideally suited to data analytics, and you shouldn’t miss out on this great opportunity to get ahead of the curve and boost your future career.

Why now? Let's have a closer look in our upcoming blog entry. Follow-up here.


From this series:

Part 2 - Engineers Make The Best Data Scientists!

Part 2 - Engineers Make The Best Data Scientists!

 

Previous
Previous

Part 2 - Engineers Make The Best Data Scientists!

Next
Next

NQI - One Of Many DOCSIS Health Indicators For Cable Modems In HFC Networks