In this article, we’ve outlined the benefits of establishing an in-house machine learning team: which companies should consider this approach, which options they could pick, and which pitfalls they may face on this journey.

SwissCognitive Guest Blogger: Dmitry Bubnov, CEO and co-founder of ENBISYS

SwissCognitive, AI, Artificial Intelligence, Bots, CDO, CIO, CI, Cognitive Computing, Deep Learning, IoT, Machine Learning, NLP, Robot, Virtual reality, learningWhich factors should companies keep in mind before hiring a team of engineers for an inside Machine Learning project?

Machine Learning is a progressive tool for extracting correlations, analyzing hidden patterns, and learning from data. A proprietary cloud-based machine learning solution can help businesses identify trends and provide insights into business processes, finances, offerings, and customers. It can also be used to automate some parts of day-to-day activities.

That is why many companies are currently experimenting with various strategies, technologies, and opportunities to capture value in the growing market of Artificial Intelligence (AI) applications, and all of them require significant investments.

In this article, we’ve outlined the benefits of establishing an in-house machine learning team: which companies should consider this approach, which options they could pick, and which pitfalls they may face on this journey.

Why set up an in-house Machine Learning team

If you are thinking of hiring your Machine Learning team, your organization must be having a lot of practical problems that cannot be solved by using the means available. Using data analytics is probably long overdue because you’ve been collecting data for years, and you have well-established policies and procedures for data collection. It would also be great if your experts are already making an effort to solve those problems based on these data, but they naturally lack skills in data analytics. If this is the case, there is a high chance you can  create and train an ML model to solve these tasks more efficiently than humans.

As an example, think of the operational processes of an oil-producing company dealing with the issue of annular pressure exerted by the hydrocarbons in the annulus of a well or between two strings of casing inside a producing well. Such pressure needs to be handled by a dedicated crew to avoid any workover operations. A machine learning model could be used to predict the time the crew should start operations up with the accuracy up to +/- 1 hour, instead of +/- 5 hours like traditional tools offer. This way, you can better plan the workload of highly qualified personnel and reduce the associated costs by up to 30%. And there is no way you can purchase a box solution available in the market to solve this issue.

Which companies should consider an in-house Machine Learning team. 

  • Large scale. If you are a small or medium organization, think of working with an external ML team instead. It would be a safer option to do the research and test your hypotheses with a 3rd party provider before you commit to a long-term project and invest into an in-house Machine Learning Team.
  • Enough challenges. There must be a lot of problems to be solved for several years ahead, and the new data should keep coming in. Otherwise, your highly qualified and professional team will not feel challenged enough and you will face issues with employee retention.
  • You must have valid internal policies for data collection, security, and communication between teams and departments. If you don’t have those in place, this will slow down the research process. Will your data scientists even be able to bypass the security policy and obtain the required data?
  • You should be ready to invest around $35-40K a month to keep a team of 3-5 Data Scientists.


Outsourcing ML services might be a good idea if

  • You haven’t yet proven that your data-based approach is going to work for your exact business and your specific challenges.
  • You are not sure you are ready to integrate a new data-centric culture, where all the teams are on board with this innovation, or change your security policies to make the data available to the new team.
  • You are not sure there will be enough problems to solve. If your new team makes one or two ML models and the specialists are then kept just to maintain these models, they are likely to get bored and look for employment elsewhere.


The options for establishing an in-house Machine Learning Department outlined below are intended for the companies who have made a managerial decision to have an in-house team instead of working with an external provider. Each of them has its benefits and disadvantages, so you can pick the option that is best suited for your situation.

1) Grow the team of machine learning engineers

Growing your team of Machine Learning engineers means giving ML tasks to existing members of your IT department and letting them learn by solving some mock (made up) challenges.  While this way is relatively safe, it will require a long time and high costs. Dealing with less experienced engineers will not let you build a solid R&D methodology, require too much CTO engagement, and lead to incorrect task setting. This may result in long project timelines and low quality of the overall project outcome.

Thank you for reading this post, don't forget to subscribe to our AI NAVIGATOR!


“Machine Learning Competencies Range. Option 1.“

„Machine Learning Team Setup Scenario 1“

2) Hire experienced data scientists

Hiring 2-3 experienced data scientists for your existing commercial project will mean a fast-established team. The main concern is: will you be able to speak their language and set their tasks properly? If the answer is no, you might end up setting wrong expectations, incurring high costs only to neglect your major project and fail the new one. However, this approach might work if your CTO has at least some experience managing ML R&D process.

„Machine Learning Competencies Range. Option 2“

“Machine Learning Team Setup. Scenario 2”

3) Learn through courses available

When you have your existing team members take a couple of courses on Machine Learning, you may end up with an exciting outcome. You will have a solid R&D methodology along with good theoretical knowledge of the modern technology stacks available. However, it will probably be difficult to apply the acquired theory to real-life projects as well as come up with a clear strategy for ML projects.

„Machine Learning Competencies Range. Option 3“

„Machine Learning Team Setup Scenario 3“

4) Hire CDO/CDS

Hiring a Chief Data Officer or Chief Data Scientist is a great option that guarantees project success. Your team will be established within a short term and there will be a minimal risk for the project. The only problem is that it is almost impossible to source such specialists in the labor market, and their skills and competencies are very hard to assess. You might end up hiring the wrong guy for your project while the cost will be very high.

„Machine Learning Competencies Range. Option 4“

„Machine Learning Team Setup Scenario 4“

5) Dedicate/Hire junior Data Scientists and hire ML consultants

The most successful case would be assigning a part of your team or hiring junior Data Scientists and hiring an experienced consultant to set up the processes to make sure your business challenges are met. This will speed Option 1 dramatically by putting in place the code style, the algorithms, the frameworks, approach, and R&D methodology. As a result, you will have solid ML competencies in-house together with the right data strategy and right specialists to implement it.

„Machine Learning Team Setup. Scenario 5“


When your organization needs a Machine Learning team, you should first estimate your chances for success. Do you have a data culture in place? Will there be enough tasks to keep the team busy for years to come?

Anyway, it’s always good to start a pilot project with an external provider to establish the frameworks, communications and minimize risks. This will help you reveal, specify, and motivate the concrete project goals and have the first tangible outcome. Once the pilot is complete, you may start looking for ways to hire in-house engineers keeping in mind the options that we’ve listed.

About The Author:

Dmitry Bubnov is an experienced technology leader with a 15-year record in custom software development for Enterprise projects in eHealth and EdTech. Dmitry has strong competences in forming and managing highly-efficient dedicated development teams with a focus on AI machine learning, web applications and cloud solutions.