Automatic learning and ai are among the most popular themes today, especially within the technological space. I am lucky to work and develop with these technologies every day as an automatic learning engineer!
In this article, I will guide it through my trip to become an automatic learning engineer, throwing some light and advice on how you can become one yourself!
My background
In one of my previous articles, I wrote extensively about my trip from school to ensure my first data science work. I recommend you Look that articleBut I will summarize the key timeline here.
Almost all in my family studied some type of Stem subject. My great -grandfather was an engineer, my two grandparents studied physics, and my mother is a math teacher.
So, my way was always paved for me.
I chose to study physics at the university after seeing the Big Bang theory at age 12; It is fair to say that everyone was very proud!
At school, I was not dumb in any way. It was actually relatively bright, but I did not apply completely. I obtain decent grades, but definitely not of what was completely capable.
I was very arrogant and I thought I would do well with zero work.
I asked the best universities such as Oxford and Imperial College, but given my work ethics, I was delirious thinking that I had a chance. The day of the results, I ended up in cleaning when I lost my offers. This was probably one of the saddest days of my life.
Cleaning in the United Kingdom is where universities offer places to students in certain courses where they have space. It is mainly for students who do not have a university offer.
I was lucky that they offered me the opportunity to study physics at the University of Surrey, and I got a first class master's degree in physics!
Genuinely there is no substitute for hard work. It is a crying cliché, but it's true!
My original plan was to do a doctorate and be a full -time researcher or professor, but during my title, I did a year of research, and I felt that a career in research was not for me. Everything moved very slowly, and it didn't seem that there were many opportunities in space.
During this time, Deepmind launched its Alphago – The movie Documentary on YouTube, which appeared in my home feed.
From the video, I began to understand how I worked and learned about neuronal networks, learning reinforcement and learning. To be honest, to this day I am not yet an expert in these areas.
Naturally, Cavé dear and discovered that a data scientist uses ai algorithms and automatic learning to solve problems. I immediately wanted to enter and began to request graduate roles of data sciences.
I spent countless hours encoding, taking courses and working on projects. I requested a 300+ jobs And finally landed my first postgraduate science science scheme in September 2021.
You can listen more about my trip from a podcast.
Data science trip
I started my career in an insurance company, where I built several supervised learning models, mainly using gradient -driven tree packages such as Catboost, XGBOOST and Generalized linear models (GLM).
I built models to predict:
- Fraud – Does anyone fraudulently claim profits?
- Risk prices – What is the cousin that we should give to someone?
- Number of claims – How many statements will someone have?
- Average claim cost– What is the average claim value that someone will have?
I made about six models that cover the regression and classification space. I learned a lot here, especially in statistics, since I worked closely with the actuaries, so my mathematical knowledge was excellent.
However, due to the structure and configuration of the company, it was difficult for my models to move beyond the PIC stage, so I felt that I lacked the “technological” side of my set of tools and the understanding of how the Companies use automatic learning in production.
After a year, my previous employer communicated with me asking me if I wanted to request a role as a junior data scientist who specializes in Time series forecast and improvement issues. I really liked the company, and after some interviews, they offered me the job!
I worked in this company for about 2.5 years, where I became an expert in combinatorial prognosis and optimization problems.
I developed many algorithms and implemented my models for production through AWS using the best software engineering practices, such as unit tests, a lower environment, shadow system, CI/CD pipes and much more.
Just say that I learned a lot.
I worked very closely with software engineers, so I collected many engineering knowledge and continuous automatic learning and self -study statistics.
I even He obtained a promotion From the junior level to the medium at that time!
Transition to mle
Over time, I realized that the real value of data science is using it to make live decisions. There is a good appointment of Pau Labarta Bajo
ML models within the Jupyter notebooks have a commercial value of $ 0
It makes no sense to build a really complex and sophisticated model if it will not produce results. Looking for additional 0.1% to reach multiple models is often worth it.
It is better to build something simple that can implement, and that will bring a real financial benefit to the company.
With this in mind, I began to think about the future of data science. In my head, there are two ways:
- Analytics-> You work mainly to obtain information about what the business should be doing and what you should be considering to increase your performance.
- Engineering-> You send solutions (models, decision algorithms, etc.) that provide commercial value.
I feel that the data scientist who analyzes and builds few models will be extinguished in the coming years because, as we said previously, they do not provide a tangible value to a business.
That does not mean that they are completely useless; You must think about it from the commercial perspective of your return on investment. Ideally, the value it contributes must be more than its salary.
You want to say that you made “x that it produced and”, that the two previous ways allow you to do.
The engineering side was the most interesting and pleasant for me. I really enjoy encoding and building things that benefit people, and that they can use, so naturally, that is where I gravitated.
To move next to ML engineering, I asked my line manager if I could implement the algorithms and ML models that I was building. I would get help from software engineers, but I would write the entire production code, make my own system design and configure the implementation process independently.
And that is exactly what I did.
Basically I became an automatic learning engineer. I was developing my algorithms and then sending them to production.
I also took Data Structures and Neetcode Data and Algorithms Course To improve my foundations of computer science and start Bloguear about software engineering concepts.
Coincidentally, my current employer contacted me at this time and asked me if I wanted to request a role as automatic learning engineer who specializes in general ML and optimization in his company.
Call it luck, but clearly, the universe was telling me something. After several rounds of interviews, they offered me the role, and now I am a full automatic learning engineer!
Fortunately, a type of “Fell To Me” role, but I created my own luck through the ability and documentation of my learning. That is why I always tell people to show their work: you don't know what it can come from it.
My advice
I want to share the main advice that helped me make the transition from an automatic learning engineer to a data scientist.
- Experience– An automatic learning engineer is No an entry level position in my opinion. It must be well versed in data science, automatic learning, software engineering, etc. He does not need to be an expert in all of them, but having good foundations in all areas. That is why I recommend having a couple of years of experience as a software engineer or data scientist and other areas.
- Production code-If it is from Data Science, you must learn to write a good and well proven production code. You should know things like writing, fluff, unit tests, formatting, mockery and ci/cd. It is not too difficult, but it only requires some practice. I recommend asking your current company to work with software engineers to get this knowledge, it worked for me!
- Cloud systems– Most companies today implement many of their architecture and cloud systems, and automatic learning models are no exception. Therefore, it is better to practice with these tools and understand how the models are activated. I learned most of this at work, to be honest, but there are courses that you can take.
- Command line– I am sure that most of you already know it, but every technology professional must be competent in the command line. It will use it widely when implementing and writing production code. I have a basic guide that you can pay here.
- Data and algorithms structures– Understanding the fundamental algorithms in computer science are very useful for mle roles. Mainly because it is likely to be asked about it in the interviews. It is not too difficult to learn compared to automatic learning; It only takes time. Any course will do the trick.
- Git and Github– Once again, most technology professionals should know GIT, but as MLE, it is essential. How to crush the commitments, make code reviews and write pending extraction applications are essential.
- Specialize– Many mle roles that I saw required that you had specialization in a particular area. I specialize in the weather forecast, optimization and general ML based on my previous experience. This helps him highlight in the market, and most companies are looking for specialists today.
The main theme here is that I basically described my software engineering skills. This makes sense since it already had all mathematics, statistics and automatic learning knowledge of being a data scientist.
If I were a software engineer, the transition would probably be the opposite. That is why ensuring a role as automatic learning engineer can be quite challenging, since it requires competition in a wide range of skills.
Summary and more thoughts
I have a free bulletin, Distribute the datawhere I share weekly advice and advice as a scientist of exercise data. In addition, when you subscribe, you will get my Free Data Science Curriculumand Short pdf version of my road map ai!