Initiating a machine learning project: the skills your company needs
Last week, we explained what machine learning is, how it works and how big this topic is nowadays. Before our free annual event Innovation Leader 2018 – Machine Learning Business Implications on April 26th, we would like to give you an insight into the “must have” skills your company needs to launch a machine learning project.
Get used to unclear boundaries
Machine learning is a relatively new field, very exploratory and experimental. It can be difficult to spot the right skills and to define clear boundaries between the roles of the different actors who come into play in a machine learning project. There is a myriad of roles and technologies used by experts within the area. We encounter so many different titles and descriptions (which by the way differ from one firm to another) that it is often difficult to spot who/what is needed. Moreover, the composition of a machine learning team will change in response to the nature and the timing of the project.
As you initiate a machine learning project, you first need to define the business problem you are looking to solve and whether your firm has the right data to achieve it. At this stage, you will need a data scientist. Below you will find the skill set of a modern data scientist.
Source: Modern Data Scientist skill set by Marketing Distillery
Be careful though. There are several different types of data science specialists and it can sometimes be difficult to spot the kind of data scientist your firm/project needs. In the context of machine learning, there are mainly three profiles that will be the most useful.
Data analysts: they collect data, clean it and look for patterns in the data or model it. These insight-type people help you in the early stages of your machine learning project (data preparation) by identifying and understanding the trends in your data and determining any problems that must be corrected. Data analysts are not “product people”: they are not directly involved in the production experience (what the end-user will see).
Data strategists are essentially acting as business owners for data. They understand the business side of the equation but also know what the data analysts are doing and which tools they are using.
They have the necessary expertise to understand the identified problems and they will make sure your firm has the right data to solve them. They will also develop business use cases.
With a strong background in software development and engineering, machine learning engineers will build machine learning solutions to solve business and customer problems. Their role is to deploy models, manage infrastructure and run operations related to machine learning projects.
Regardless of their specialization, all data scientists have applied statistics skills and can derive meaning from large data sets using statistical models. What you will want to look for though are data scientists who combine deep technical knowledge with enough expertise in your domain to address your business needs.
Data scientists should be able to steer machine learning projects. They will be helped by software and data engineers. Indeed, as machine learning projects require vast amounts of data and computation power, data engineers will deal with the infrastructure aspect and solve the challenges raised by large data sets.
Data analysts, data strategists, machine learning engineers and data engineers will be the core team of your machine learning project. The image on the left will help you visualize the different competencies they have.
Depending on the nature and timing of your project, you might also have a look at other profiles.
Projects in fundamental research will require researchers/research scientists. Research scientists will focus on undeveloped and untested new technologies. They will often work on promising data leads uncovered by data scientists. Research scientists are focused on driving scientific discovery and are less concerned by industrial applications of their findings.
If your project is closer to production, the applied research scientists/engineers’ profile will be of interest for you. Applied research scientists/applied research engineers straddle engineering development and research and will be more concerned with practical research such as identifying and implementing industrial applications for scientific discoveries.
A user experience and user interface designer (UX/UI designer) is also a profile you want to gather for your machine learning projects. This person will deal with how the product is laid out and how the product feels.
Finally, a product owner or manager will round out your machine learning project team keeping in mind questions like “does the machine learning fit the product goal?” or “what interactions, actions and control do users have?”
Now you know and can differentiate the roles and skills that might be involved in a machine learning project. You have to expect, though, that the practice will differ from the theory. Indeed, due to the constant adaptation required by the machine learning field, the responsibilities of the positions can overlap.
Keep an eye on certain characteristics
You don’t manage a machine learning project like you manage a software development one. Even if a machine learning project does need the usual skills required for a software development project, a software engineer doesn’t have the same working environment as a machine learning professional. The software engineer will usually jump from one project to another having always structured tasks and a clear delivery timing. The day-to-day work of a machine learning professional is made up of exploration and experiments with less clear timelines. Therefore, you will have to keep an eye on the essential characteristics to look for in a machine learning project’s team member.
Beside a solid background in mathematics and statistics which is crucial to understand which algorithms best address a problem and how to interpret and optimize outcomes, your machine learning project will need people who have the capacity to learn and adapt fast. Machine learning is developing at a fast pace and keeping up is critical. New algorithms involve new strategies. Therefore, your machine learning team members have to be up-to-date and to adapt constantly.
Going beyond understanding data will also be a must. Indeed, the purpose of the analysis and the applied implications of the research are what really matters.
As tools and methodology in machine learning are relatively new, you will also need highly creative team members. Your firm needs people who are able to come up with creative ways to tackle a problem.
You must clearly define what is that you want from machine learning to initiate a project in that field. The questions “What type of problem my firm needs to solve” and “how quickly must the problem be solved” constitute the starting points of your reflexion related to the needed skills for such projects. You will then look for the right people knowing that data scientists will be the core experts of your machine learning project team.