Marcello Politi

Photo by Pawel Czerwinski on Unsplash Learn how to use the Template design pattern to enhance your code Marcello Politi · Follow Published in Towards Data Science · 4 min read · 12 hours ago — Introduction Recently I’ve been working on the domain-specific fine-tuning of several LLMs. The first and maybe the most important part of this task is to collect, scrape, and clean textual data to feed the LLM. I noticed that my

Photo by Sonika Agarwal on Unsplash Learn how to use queues, daemon threads, and events in a Machine Learning project Marcello Politi · Follow Published in Towards Data Science · 7 min read · 10 hours ago — Introduction In most Machine Learning jobs, you won’t do research on improving some model architecture or designing a new loss function. Most of the time you must utilize what already exists and adapt it to your use

Photo by Milad Fakurian on Unsplash Learn how to implement coding best practices to avoid tech debts Marcello Politi · Follow Published in Towards Data Science · 5 min read · 7 hours ago — Introduction Many people approaching machine learning don’t have a strong background in computer engineering, and when they need to work on a real product their code can be messy and difficult to manage. This is why I always strongly recommend

Photo by Simon Wiedensohler on Unsplash Quickly set up LLM APIs with BentoML and Runpod Marcello Politi · Follow Published in Towards Data Science · 6 min read · 2 hours ago — Introduction I often see data scientists getting interested in the development of LLMs in terms of model architecture, training techniques or data collection. However, I have noticed that many times, outside the theoretical aspect, in many people have problems in serving these

Photo by Michael Dziedzic on Unsplash Run deterministic and non-deterministic tests to validate your dataset Marcello Politi · Follow Published in Towards Data Science · 9 min read · 7 hours ago — Introduction In an MLOps pipeline where we try to automate as many steps as possible, keeping in mind the goal of minimizing the number of errors that direct intervention by a programmer might cause, it is also important to take care of

Photo by Leo_Visions on Unsplash Learn how to build an efficient pipeline with Hydra and MLflow Marcello Politi · Follow Published in Towards Data Science · 7 min read · 6 hours ago — Introduction When we develop Machine Learning models, we usually need to run lots of experiments to figure out which hyperparameter setting is best for a given algorithm. This can often lead to dirty code and losing track of which result…