Summer is over! Time to MLOps

This summer was rough.

The beguining was going well. I finished my project for Predictive Methods of Data Mining at NOVA IMS. I’m pretty happy with the results. We ended up used streamlit, and we got a pretty good score on Kaggle with our rather simple neural network. =D
I also watched the LLM Bootcamp 2023, which gave me a lot of insight on the current state of LLMs.

But my side projects were slowing down. I guess I was able to picture a structure in my head that might work (front end, backend’s data flow, data storage, RAG adn LLM inteerface) but I was having trouble implementing it.

But what threw me off was a course I attended: Healthcare NLP for Data Scientists on July 19-20, 2023 from John Snow Labs. Don’t get me wrong, the course was great, I will need to revisit for sure, but it made me realize that I was lacking solid foundations ML, NLP and as a data scientist in general.

It took me the rest of the summer to figure out what to do next.

And to gather motivation to do it.

I realized that MLOps might be the solution I was looking for, thanks to this video from Noah Gift I ended up enrolling in the Coursera’s MLOps | Machine Learning Operations Specialization. I’m currently finishing the 2rd course. Pretty good so far!

DevOps concepts are helping to be more consistent and efficient in my work.

I’m more agile working with the command line and github.

I’m dipping my toes in Docker, AWS and Rust.

I’m even starting to realize that I might be need to create a python library for my data visualization projects, so i can replicate with ease the reports i’ve been doing for my clinical context.

Let’s push through the rest of the year!