19 nov 2018
- Why is it so hard to hire AI and ML talent?
- Where to find AI and ML talent?
- What skills to look for in AI and ML engineer?
In recent years, machine learning algorithms have become more accurate, and many companies use them as a competitive advantage. In fact, almost every company on Fortune 500 list invests in machine learning to run more effectively, and eventually ‒ to drive more sales. To top it off, all big companies today are using machine learning behind the scenes, and they have already seen a positive ROI.
Why is it so hard to hire for this position? On the other hand, hiring a machine learning or AI talent comes with a lot of challenges. On average, a machine learning engineer earns 115,000 USD a year in the USA. Even if you’re the lucky one and this digit is within your budget, the talent pool often appears to be shallow.
Creating machine algorithms differs from developing a common app – fewer people are qualified to build viable artificial intelligence solutions. To make the matters worse, nearly 80% of machine learning engineers are scrambled by Fortune 500 companies. Companies like Google, Amazon and Microsoft spend as much as $650 annually to retain their AI experts.
Why should you push your search outside of your region?
While in North America you will probably face the shortage of talent, Eastern European talent pool, and Ukrainian in particular, might have something to offer. The first benefit is financial – the average salary of data scientist is $25,000 (2017), much lower compared to the USA. Ukraine boasts several world-recognized companies that use AI and data science to create their own products. The brightest one is Grammarly – a tool that uses Natural Language Processing to check texts for grammar and plagiarism. Other examples are Scorto, Samsung R&D Institute, gaming companies Vostok Games and GSC Game World (the guys who created the famous STALKER).
How to recruit a good data science and machine learning specialist?
Designing AI solutions requires a strong understanding of mathematics, statistics, background in data science and programming combined with a great deal of professional intuition.
NCube, a software company that builds and hosts tech teams in Eastern Europe, gives recommendations regarding the candidates’ pre-selecting process.
- Theory scoring
Check the candidate’s understanding of linear algebra, statistics, Bayes Nets, Markov Decision Processes, probability theory and techniques, differential and integral calculus, probability distributions as Machine Learning is closely related to these disciplines.
- Programming challenge
Apart from data structure, OOP and algorithms knowledge, an ML engineer masters at least two programming language with machine learning capabilities. The most popular ones are C/C++, R, and Python.
- Data engineering skills
The role of a machine learning engineer requires working with huge amounts of data – analysis, processing and data visualization. As such, knowledge of databases such as SQL, NoSQL, ETL among others is required.
- ML libraries & frameworks
Understanding of static and dynamic libraries and algorithms, their assets and weaknesses, and how they apply to each technology: TensorFlow, Azure, Microsoft CNTK, Theano, Caffe, Keras, Torch, Spark MLib.
Machine learning specialists deal with linear regression, local and kernel regression, neural networks, and reinforcement learning.
- Software design
Algorithms are usually integrated into crucial software components, so it is advisable to test a candidate’s comprehension of programming interfaces. It would be a bonus if a candidate has hands-on experience in interface design.
When hiring machine learning talent to create novel algorithms, prepare to undertake a challenge of estimating their technical skills. These include theoretical knowledge, mastering at least two programming language and a few libraries and frameworks, together with a solid understanding of software architecture.
The best weapon in the war for AI and ML talent is high salaries. However, should you want to build your nearshore virtual development team (which is a more budget-friendly strategy), NCube can be your trusted provider. If you want to build a 100% PhD team, we can do it, too.