Hi everyone! @radek: thanks a lot for starting this thread. I think that’s a very important issue to discuss. @ everyone else: thanks for sharing your stories. I decided to share mine as well.
I fall exactly into the problematic category that Radek mentioned. I’m neither a domain-expert learning to code nor a software engineer getting practical ML experience. During my master’s degree in Economics I programmed for the first time in my life at age 25 (very basic statistical analyses in R). After graduating I started to work as a project manager in Business Intelligence and while I liked the job, I really missed the practical experience of building stuff myself. So I set out to learn Python in Datacamp and did some (rather unnecessary) ML courses on Udemy. A big game changer for me was taking the deeplearning.ai specialization, which really got me hooked on deep learning. However, when I tried to start my first project after the course and sat in front of an empty Jupyter notebook, I really didn’t know where and how to start which made me question the whole self-education thing and got me to the point of stopping. I’m glad I didn’t give up though, because then I decided to check out fastai and that’s when things started to change for me.
At this point I had already been talking to my boss for a while about changing my position and after some time he agreed to start a new project that involves NLP and ML where I would be in charge of the entire data science work. So parallely with taking the fastai DL course part 1 I started applying the
very same methods in my own work project. Because I was the entire “Data Science team” in the company, my learning curve was incredibly steep and hard to climb, but after all the project turned into a success.
I also want to briefly talk about locations. I live in Prague, Czech Republic, so I am not located in an international tech hotspot like San Francisco or London. I would be open to remote work, but I prefer to work locally. I think it can even be an advantage for outsiders to work in places other than the well known hotspots. While there aren’t that many job offers available, also the competition on the labor market is much lower. And since more and more traditional companies are willing to try out AI, such places might be good for outsiders to get into the field.
Of course this way won’t be possible for everyone and I also consider myself very lucky, but I wanted to share my story because it may show another option for folks who want to land their first data science job. If you have a good relationship with your boss, you can try to convince him/her to let you try out a project that involves ML (maybe part-time or even on your own time). Choose a small project that is likely to succeed because you most likely will only have one shot of convincing your boss.
Thanks @radek bringing me to this interest thread.
I barely know any programming when I graduated from college, I have only took 1 c++ course.
Before I start my first job, I spend 3 months to finish Andrew Ng machine learning course and Data analyst nanodegree from Udaicity. These are very useful resource at the time, as there are not much resource at the time, Udacity provide a structure that I can begin with.
Not to mention there was 0 degree about data science in my country at the time, so online resource is the only way. I got a BI related job, and get an opportunity to transfer to data science team after one and half year. I start fast.ai in 2017, and follow along until fastai v3.
From my experience, online courses are unlike college degree that can be put on resume, but the skills learnt are very useful. I cannot compare it to college studies since I do not have much experience. I think my skills is taking off at 2018, when I started to make some PRs to fast.ai (they are not significant, but the process of learning how to dig into a code base, getting familiar with linux or cloud, debugging is very useful, arguably much more important than knowledge about algorithm in real life.)
I am curious if people find works remotely or freelancing.
Wow your story resonate with me! We have a similar path…Managed BI hardware/software project development but no practical coding experience. I took Data Camp coding class in Python and found a job as a programmer in the medical field, but no AI yet.
Right now, my day to day is very hard. I’m not in my DataCamp sand box anymore. I have to learn shell, Bash, JavaScript, Networking, Data base, etc… On top, I have to learn the medical slang. The first 6 months were awful
What do you recommend to move toward AI/ML? Build a foundation in DevOps for a few years or try to switch as soon as I can? Was it worth it ? What don’t you like working in AI ?
I am sure that your programming skills are a major asset when switching to AI/ML. If you really want to get into the field, I would say don’t wait and give it a shot. In case you haven’t taken the fastai courses yet, of course I highly recommend doing that.
I didn’t switch from a programmer job into ML/AI, so I cannot comment on that. But I am quite happy with my decision of switching from project management to data science. A challenging part of working in AI, in my opinion, is the fact that you usually don’t know in advance what is going to work. And if your approach isn’t working, often it’s not obvious how to improve your solution.
To answer your question about taking the class…Yes I did listen to part 1 in fastai V2 in 2017-2018 (videos 1 to 7) and successfully install it on my local machine (which was a pain in the b**!! and took so much time). Then I sat on it a while. Kinda of discourage a bit…and got away from it.
When the new fastai version came out, I bought the book and open a Paperspace account using the their free server(Really worth it!! You save so much time) and I was able to run personal experiments with the vision module.
But, I haven’t done any "production "projects with it and surely don’t master it yet.
First of all thanks @radek for starting this topic. I discovered this post quite late and there might be a chance that this post is no longer that active, but I will post here anyhow.
I am software test Automation engineer based out of Pune, India. I started my career as a Test engineer (no coding experience). Three years into my career I broke into coding (using python) and witched career into Test automation. Then three years after that got interested in Machine learning/deep learning and went through numerous articles, tutorials etc. etc. from both online and offline sources in the hope of learning about this new field, back then ML MOOC, tutorials courses were math heavy and almost all these courses/resources used such heavy mathematical terms that I gave up on learning ML (you see I am not a math person and I am a very slow learner).
Then I came across fastai during the beginning of 2019 and came to know that if taught in a proper way ML/DL can become uncomplicated and you don’t need tons of math to learn deep learning.
After going through the courses and doing some tinkering on my own through my blog and side projects, I started doing small projects in my current place of work. I literally dug our use cases where I could apply simple models to do simple tasks. Even I became ML evangelist for my organization.
By now I had implemented some Deep learning-based tools at my place of work and have done a few projects through my blogs. Since the past 1 year I am trying to switch career to Machine learning.
In India we have lots of startups and companies where Deep learning-based tools/projects are in widespread use. However, there is a lot of gatekeeping across industries here when you try to get into machine learning from some other domain. It becomes hard if you are not a “full stack developer” by profession or an “experienced machine learning engineer” (the magic number is 5yrs. experience into ML).
I would like to know if this is the experience with everyone from a non-ML background here? and what would be some suggestions to change career into a DS role if you don’t have prior experience of being a ML engineer or a Data scientist.
From what I understand there is no clear path in ML. You have to find your own path. If you like it and apply it even at small scale you are already ahead of most people around you. I recently discussed with my old friends from college (all engineers) and I asked them if they would have problems to solve with ML/AI most of them told me no or what is ML/AI.
We are still very early in ML/AI history. Most people don’t know what do to with it. So nobody see a use for it. Hope it answers some of your questions.
Thanks for your comment. You are right that we are in the early history of AI/ML and the best thing to do is to keep going by finding projects to work on.
However, the other part of the truth is that it’s difficult to get into a ML role unless you have the “professional” experience un ML/DL rhat you could show in your peofile. I feel that in today’s time the firms which accept people into this field irrespective of any prior experience in ML,DL,DS would be the place from where innovation would come out.
