Paulo Sampaio

Senior data scientist with 10+ years of experience, delivering AI solutions across diverse industries.

Projects

This is a selected collection of projects I've worked on — a mix of professional work, PhD research, teaching examples, and personal explorations. It's not everything I've built, but it's a set that I feel represents the kind of work I enjoy: practical, diverse, thoughtful, and creative.

Project thumbnail

LLM based medical report parser

Use LLM to extract structured output from free text reports

Project thumbnail

LLM product description generator

LLM to automate the creation of product description tailored for your audience

Project thumbnail

Mask detection and alert system

End-to-end pipeline for image detection, classification and alarm system

Project thumbnail

Medical image segmentation

Custom UNet and CNN for medical image segmentation and classification

Thumbnail of Astro arches.

LLM based text-to-image and image-to-image search

Based on text or image, find closest product on database.

Project thumbnail

Spotify music classifier

Teaching supervised learning using spotify playlist

Work Experience

•
PhD student / AI Researcher

Universität Bern - ARTORG Center
2021 - Present

PhD in Biomedical Engineering, focusing on Artificial Intelligence in medical imaging. I develop computer vision algorithms for pancreatic tumor segmentation and classification. I also deploy a MedGemma LLM for automatic extraction of structured information from medical reports.
•
Analytics manager

JoGo Growth experts
2020 - 2021

Provided analytics consulting services for tech startups in Brazil. Analytics lead for a prop-tech unicorn, automated data-driven sales and pricing decisions for a retail-tech marketplace
•
Data science specialist

McKinsey & Company
2018 - 2020

Responsible for analytics workstream on projects in Latin America. Implemented end-to-end solutions using gradient boosting, MLP, CNNs, prophet (time series). Built and deployed ETL pipelines using tools such as Kedro, metaflow, and prefect to connect and process different sources of structured and unstructured data to machine-learning applications
•
Data science tech lead

EDITED
2014 - 2018

Trained multiple CNNs for identifying colour, pattern, shape and detail in product images, including feature vector extraction, vector database and fast vector matching and retrieval for image-based product matching similarity search. Built NLP model to classify e-commerce products into internal taxonomy.
•
Data scientist (MSc internship)

Accenture
2013 - 2014

Worked with R and SQL for big data handling, machine learning and data analysis. Contributed to projects such as: predicting failures in wind farms, forecast and visualization of wind direction/intensity, confidence of car's engine warning system
•
Revenue management sr. manager

LATAM Airlines
2007 - 2013

Managed a team of 16, controlling all strategy and business development related to the European and Pacific markets, implementing market segmentation, pricing definition, demand forecasting and inventory management.