Data Scientist | AI Engineer
I am a professional with a generalist profile, holding a Master's degree in Data Science from the University of Mannheim. I have +6 years of full-time work experience in Brazil and Germany, combining Machine Learning, Analytics, and AI techniques to transform databases into products and insights. I am currently based in Frankfurt am Main, Germany.
- Machine Learning & AI: NLP, Deep Learning, Artificial Intelligence, PyTorch, TensorFlow, RAG, Knowledge Graphs, Ontology Engineering.
- Data Engineering & MLOps: Python, SQL, ETL, Apache Airflow, Google Cloud Platform, Azure, CI/CD(/CT), MLOps, Git.
- Analytics & BI: Data Analysis, Metabase, Power BI, Excel, R, SPARQL.
Frankfurt am Main (03/2024 - Present)
- Combined LLMs and analytics into an AI assistant based in RAG and text-to-SQL for self-service BI that answers natural language questions about the company KPIs with data and citable sources (93% accuracy).
- Built the company's product demand forecasting model using deep learning algorithms (5% relative MAE) and following MLOps best practices.
- Modelled customer segmentation, churn, and product recommendation with explainable Machine Learning for tailored Marketing approaches.
Remote (05/2021 - 12/2023)
- Created a RAG pipeline to generate grounded legal commentaries about thousands of Brazilian law articles using GPT-4 on public Brazilian court decisions (~0.8 BERTscore).
- Developed a text classification model to detect relevant court decisions regarding Data Protection with 0.9 F1-score using NLP transformer algorithms.
- Built the company's churn prediction model using Machine Learning algorithms, scaling individual Customer LTV estimations with 0.85 F1-score.
- Created +20 KPI-tracking dashboards for Product, Customer Service, Sales, Marketing, and Finance teams.
Data Analyst & BI Intern @ Cogna Educação (08/2018 - 05/2021)
- Implemented Saraiva Educação's book recommender API system, using Python, Machine Learning and cloud functions.
- Constructed +10 ETL pipelines to ingest data into BigQuery and +100 Metabase dashboards for Product, Marketing, CS and Sales teams.
- Automated the generation of performance reports sent to the company's B2B clients using Python Django.
Graduated in 2024
- Thesis: Binarization of Knowledge Graph Embeddings with Semantic Preservation - awarded as best thesis by the School of Business Informatics (1,0).
- Seminar: Semantic Change and Time-Awareness of LLMs.
- Projects: Improving Interaction in Knowledge Graphs with Embeddings; Node Classification of Scientific Papers using GNNs; Info Retrieval & Web Search MS MARCO; DBpedia-based Akinator using SPARQL queries.
Graduated in 2019
- Degree focus: Thermodynamics, Unit Operations, Kinetics, and Reactor Design (GPA 82%).
- Extracurriculars: Scientific Research (Nanotechnologies and Contaminant Adsorption), Junior Consulting, social volunteer work.
- Exchange programs: Friedrich-Alexander Universität Erlangen Nürnberg and RWTH Aachen University.
- Knowledge Graph Engineer (metaphacts, 2023)
- Data Science & Machine Learning (Tera, 2021)
Wikinator: an Akinator-like game based on DBPedia's Knowledge Graph
Team project for the module of Semantic Web Technologies / Knowledge Graphs. It is an application inspired by Akinator that tries to guess which real world person or fictional character the players are thinking of, relying solely on data available in DBPedia's Knowledge Graph and using only SPARQL queries for data extraction.
Web Structure Mining: Paper importance prediction using Graph Neural Networks
In this work, Graph Neural Networks and traditional Machine Learning approaches were exploited to predict the importance a paper will have once it is published, given the citation network.
Information Retrieval: Learning-To-Rank with embeddings for document retrieval
Traditional IR approaches such as TF-IDF and BM-25, as well as recent embedding techniques such as GloVe, BERT and BART, were exploited in a Learning to Rank algorithm for document retrieval.
Sentiment Analysis of headlines about US presidents
The sentiment of headlines and snippets from The New York Times articles concerning two United States presidents in their first year of government were analyzed with NLP techniques.
- Portuguese: Native | English: Expert | German: Expert | Spanish: Intermediate
- Hobbies: Brewing different styles of craft beer 🍺, reading and discussing fiction in Book Clubs 📚, appreciating specialty coffee, as well as nutrition and sports (running, cycling, swimming and weight-lifting).
Feel free to reach out via LinkedIn