Coding Skills

Here is how you can evaluate my Python coding skills in detail

I have programmed in Python for more than 10 years for machine learning and AI applications. During my Master’s degree, for the ML class, I coded from scratch most ML algorithms, including a small GPT. (see data science at the beginning of my portfolio)
Please, take a look at the list of meaningful projects in ML & AI that I have assembled there for the very purpose of showing my skills in those fields.

In Comparison of Supervised algorithms, I compare all major ML algorithms on two different datasets. I coded it all up, the report alone is 10-page long with 50 plots. In the portfolio, you will also find the same analysis for unsupervised algorithms.

I completed 30 Leetcode problems, and 20 on Hackerrank, including hard ones, some with original solutions. See my Leetcode profile and my Hackerrank submissions. If you check those problems, you will learn more about my coding skills than after one hour of live coding.

In 20203-2024, I achieved 40+ certifications in ML/AI all in Python (3-4 Python projects each) from Uplimit.com, a company offering certifications prepared by industry experts.

I live and breathe by Python speed optimization. Please see my article on substack where I explain how to broadcast two meshgrids of 4 and 3 dimensions to vectorize 6 nested loops.

In 2024, I won a Hackaton in Gen AI, and placed 2nd to another one.

I have professional endorsments on Linkedin from 200+ people, some from Meta, OPenAI etc.

Finally, I currently have 4,000 lines of Python code running in a data pipeline based on AWS Lambdas (See the recommendations from the VP of engineering and the CTO).

Why I don’t do live coding tests?

It’s stressful and a waste of time because I need to re-study the things that nobody uses at work, just in case the interviewer decides to be funny (I have had unbelievable questions over the years).

I’m not just ouf of school, so many things learned in school that are not used professionally are not fresh in my mind, so, for a live coding test, I have to review all algorithms, even those no one uses at work. And I still stress because, sometimes, the interviewer asks something that is not relevant to show off. In short, it’s preparation for the wrong things.

Also I have yet to see a coding test done in the same conditions as a work environment, ie with time and all the documentation and chatGPT at our disposal. To be clear, I use chatGPT to have a first general idea, or quickly sort through all possible technical options, then I decide what to do, which library to use, the optimal data structures for the job etc. It saves a lot of time when you use a library for the first time, and very good at analyzing long error logs and traces. Of course, I don’t copy/paste its output because it would be a disaster.

Furthermore, I don’t memorize things by heart (as no one should), but conceptually, and syntax to me is just a means to an end, so I know where to go when I need to implement something. I have a library of finished projects, well organized by themes, so I quickly go to a project I know uses a particular technology to refresh my mind. Or, also, ChatGPT is great at that: I recently had to code a backend and I hadn’t used FastAPI for a while. I asked it to produce ‘code for a backend’, and I simply read the code to remind me of the syntax.

All these things are usually not possible during a live coding interview, despite being how programmers work every day, so the coding tests are not real tests of coding, but more an exercise of stress and memory.

If you still have doubts I can code, then I don’t know what to say… but … I’m happy to discuss projects I’ve done (see portfolio) for the very purpose of showing my skills, and I’m happy to do a short coding assignment if you still have doubts.

But, please, at least, first take a look at all the links I provided above.