A message to my future employer

I recently finished my BSc in Data Science and Artificial intelligence from Maastricht university (#1 AI degree in The Netherlands).

Since then, I have been working as a Software Developer for the Dutch company Sqippa, in which I developed from the ground up a full stack web application which allowed users to create their own schedules for IOT devices, including Authentication, API endpoints/requests and web deployment. (See screenshots below)

And this is not even my speciality! This project involved too much web development, which is not exactly what I am looking for to work in the future, so I decided to leave this project.

Apart from this project, I have been working on many side projects during these months, which I will explain in cronological order:

DailyClips

DailyClips was born with the idea that learning english is better via the mimic method (it basically states that a good way to learn is by hearing and repeating).

The app shows users 5 different movie clips every day and allows them to repeat what they heard, and get scored by a speech to text model.

Although the app hasn't gotten many traction, I will continue improving it as I believe is a really helpful tool.

GPT-Primavera

I wanted to see how well gpt models would perform on a set of math exams I used to participate in when I was a child: "Concurso de Primavera de Matemáticas".

These are fixed length 25 multiple choice question exams (5 options per question) that follow the IMO spirit aimed to children aged 10-18.

It turned out that, gpt4o vision model was always outperformed by the average student…

See more here

Inditex Data Science Hackathon

The hackathon was a 2 month long competition focused on making a recommendation system for a given e-commerce, in which I finally scored 17th out of 1211 competitors. In this hackathon, many participants were professional Data scientists, which made me realize I can compete even with the top professionals.

DumbGPT

Large Language Models like chatGPT have been all over the place since 2022, but few people really know how they work. Thats why I decided to learn about them following Andrej Karpathy's minGPT repository, which is the simplest implementation of GPT models.

However, I still found his repository to be too complex, therefore I created dumbGPT, which explains a full walkthrough of a GPT model from data processing until model inference.

To show it works, I applied it to a casino game and you can see the results here (In spanish):

Agent2Bench

Agent2Bench is a proposed benchmark that tests LLMs abilities in daily life computer tasks like booking flights, downloading programs or exiting vim. A demo can be found here

Jupyter Style CV Generator

A simple web application that generates CVs in the style of Jupyter Notebooks (See example below). Code here

What I've Learned

After working on all of these projects, I gained quite a lot of valuable knowledge, especially in the following aspects:

1. Technical

   For example, when starting to develop the application for Sqippa, a Senior Developer taught me how to use a browser's developer tools in order to understand how the API calls are made and reverse engineer most of the websites. That was extremely useful for the GPT-Primavera project, as instead of manually downloading the exams and solutions one by one (There were around 100 files to download) I managed to download them via code.

2. Shifted my mentality to: "Results > Process"

   During university we always had projects, however back then I was missing this critical way of thinking of: "Am I focusing on the right thing within this project?". For example: during my second year in university, we had to develop a chatbot. We had a project manual, and I remember I spent almost all the time working on the spellchecker functionality, rather than creating a cool, usable product!

3. Marketing and sales

   Six months ago, I posted this on linkedin:

   "I hate unnecessary jargon & occasional fakeness on LinkedIn. I'm just waiting to have a career that speaks for itself so my LinkedIn can look like this:"

   

   Even though I greatly have to improve this aspect, I learned that, unless you are a complete genius who doesn't need to sell themselves (and even then) it is very beneficial that you know how to sell yourself to others.

4. Knowing what I want and what I don't want

   Before finishing my degree, as most people I started actively searching for a job. I would spend hours and hours applying for jobs on linkedin, seeing offers that requested many different technologies. I managed to get a couple of interviews, however, I felt I was put into a school test with questions like: "Whats the difference between an abstract class and an interface", "Whats the difference between a tuple and a list", etc... I mean, if it's necessary I will go through that kind of process, but I had the impression that most jobs would be about working on niche super specific uninteresting parts. After working on many interesting projects, I came to realize that I am able to actually develop useful things!! 😮 And that's very important. I am now in the situation where I highly prioritize a working position where:

   1. I can learn from my team.
   2. It will have a positive impact in my career.
   3. The work I perform is impactful
   4. I enjoy it

So, if you have read all until now and would like to collaborate with me, don't hesitate to send me a message to alvaro.mrgr@gmail.com