Hamza is a junior data scientist passionnate with machine learning. In his job, he combines mathematics and computer science, for his greatest happiness.
AViSTO recruits engineers in the area of machine learning. You can find the offers on our jobs board, where it is also possible to submit a spontaneous application.
- Educational background
- Roles and responsibilities
- Supervised learning vs. Absolute Learning
- Job opportunities
Machine Learning Definition
How would you define machine learning?
Machine learning is an automatic learning process that implements a set of algorithms. These learning algorithms, for example the "artificial neural network", make it possible to teach machines to perform actions in the place of the human being.
The fields of application are very vast, since one can use machine learning from the moment when there is data. For example, at AViSTO, I worked on automated fault detection on photovoltaic panels. Or, in the area of cybersecurity, we have developed a system to detect abnormal network flows that prefigure attacks.
What educational background enabled you to work in the area of machine learning?
I did 2 years of preparatory classes, specialty Mathematics and Physics. I joined ENSIIE, a school that trains general engineers, with a focus on mathematics and computer science. In the last year, I followed in parallel a master 2 at Paris Sud University entitled "Learning, Information and Content" (Machine learning). I realized my graduation project at AViSTO, then I was hired on permanent contract.
Machine Learning Job: Roles and Responsibilities
What are your roles and responsibilities?
I give value to the data!
More concretely, I begin with an analysis phase. If we take the example of a project to build an autonomous car, I will start by looking at all available data, classify it by type (image, video, signal, text ...), extract the relevant information and determine those that will be useful for machine learning. For example, when a stop sign appears at 10 meters, it is necessary to slow down progressively and to mark a pause in front of the white line.
Then, according to the data at my disposal, I will look for the machine learning algorithm (s) that can be used. But this is only part of the problem, because it will be necessary to choose correctly the parameters of the model used: they will have a great influence on its behavior.
Then we start the learning process. We use machine learning libraries and programming languages.
What do you like about your job?
I merge mathematics and computer science!
I did a lot of maths, especially in preparatory classes: I liked it a lot, but I said to myself "what can I do with it?".
In engineering school, I discovered computer science, and I liked it too.
When, in the second year, I was told about data science, which mixes the two, it was perfect!
Machine learning skills
What skills do you need in machine learning?
There are several main skills.
I will start by talking about mathematics. They are essential first to understand machine learning algorithms, designed by mathematicians. As explained earlier, it is indeed necessary to choose the parameters of these algorithms.
That's not all, we must also have very good knowledge in probability, statistics and optimization of functions: I would say that 90% of machine learning algorithms are based on the optimization of the error function, which allows to express the margin of error of the statistical evaluations.
Linear algebra is of course also very important.
In addition to these math skills, you also need to know how to program. We mainly use R, Python or Matlab to code our algorithms.
Finally, we also use other computer languages that allow us to exploit data such as SQL or NoSQL.
Supervised learning vs. Absolute Learning
Artificial intelligence, of which machine learning is a domain, is sometimes scary ... For Elon Musk, the American entrepreneur, it could even threaten civilization.
Are you talking about the fear that machines are escaping us and taking full control?
The machine learning algorithms currently used depend on the data, we speak of supervised algorithms. In other words, the learning is limited, the machine can not do things other than those requested.
For as much, in 30 or 40 years perhaps, one can quite imagine that new algorithms make possible an absolute learning. But we are still very far away.
To conclude, how do you feel at AViSTO?
Very very good. The work atmosphere is really nice, and everyone contributes: my colleagues, my bosses, everyone participates in this positive environment. That's what struck me the most at AViSTO, and when I talk to friends who work in other companies, they confirm that it's a chance.
Machine learning jobs
AViSTO recruits in the area of machine learning! You can consult the offers on our job site, or submit a spontaneous application