João Pereira | Data Science Lead at adidas

I’m a Data Science Lead at adidas, working on the AI systems that shape how people discover and engage with products and content across adidas.com and the adidas App. I work with two teams, one focused on product discovery and one on content personalization. We use language and vision models to tag product attributes, and reinforcement learning to rank and surface content for millions of users in real time.

My work sits where research meets production: getting the science right, then building the engineering that keeps it running reliably at scale. I help shape the technical direction of our AI initiatives and guide data scientists and engineers across several locations. Before this role, I was the technical lead for demand forecasting.

My roots are in electrical and computer engineering and deep learning research, with publications cited more than 400 times. I went on to complete an Engineering Doctorate (EngD) in Data Science at Eindhoven University of Technology, building ML solutions with ASML and the Dutch Tax Authorities.

I care deeply about technical excellence, shared ownership, and teams where people are empowered to do their best work. In 2025, I was honoured to receive the adidas Superstar Award. After Lisbon, Louvain and Eindhoven, I now call Amsterdam home. Outside of work you’ll usually find me training, which I like to think counts as product research, and every now and then I write on my blog.

Career & education

adidas
2021 — Present
- Data Science Lead, Discovery & Personalization · 2024 — Present
- Technical Lead, Forecasting · 2023
- Data Scientist, Forecasting · 2021 — 2022
Jheronimus Academy of Data Science
2019 — 2021
- Data Scientist for ASML, the Dutch Tax Authorities, Van Lanschot Kempen, TE Connectivity & Heijmans
Eindhoven University of Technology
2019 — 2021
- Engineering Doctorate (EngD), Data Science
Instituto Superior Técnico, University of Lisbon
2013 — 2018
- MSc & BSc, Electrical & Computer Engineering

Highlights

All highlights

2025	Received the adidas Superstar Award
2023	Presented Probabilistic Demand Forecasting with Graph Neural Networks at ECML 2023
2023	Featured on AWS Build on Generative AI weekly series

Selected publications

All publications

Probabilistic Demand Forecasting with Graph Neural Networks

Nikita Kozodoi^*, Liza Zinovyeva^*, Simon Valentin^*, João Pereira^†, and Rodrigo Agundez^†

ECML-PKDD 2023 International Workshop on Machine Learning for Irregular Time Series, Workshop website, Sep 2023

Bib HTML PDF Poster Slides Cited by 15

@article{Kozodoi2023,
  author = {Kozodoi, Nikita and Zinovyeva, Liza and Valentin, Simon and Pereira, João and Agundez, Rodrigo},
  title = {Probabilistic Demand Forecasting with Graph Neural Networks},
  journal = {ECML-PKDD 2023 International Workshop on Machine Learning for Irregular Time Series},
  year = {2023},
  month = sep,
  url = {https://www.amazon.science/publications/probabilistic-demand-forecasting-with-graph-neural-networks},
}

Learning Representations from Healthcare Time Series Data for Unsupervised Anomaly Detection

João Pereira and Margarida Silveira

2019 IEEE International Conference on Big Data and Smart Computing (BigComp), Feb 2019

Abs DOI Bib HTML PDF Slides Cited by 124

The amount of time series data generated in Healthcare is growing very fast and so is the need for methods that can analyse these data, detect anomalies and provide meaningful insights. However, most of the data available is unlabelled and, therefore, anomaly detection in this scenario has been a great challenge for researchers and practitioners. Recently, unsupervised representation learning with deep generative models has been applied to find representations of data, without the need for big labelled datasets. Motivated by their success, we propose an unsupervised framework for anomaly detection in time series data. In our method, both representation learning and anomaly detection are fully unsupervised. In addition, the training data may contain anomalous data. We first learn representations of time series using a Variational Recurrent Autoencoder. Afterwards, based on those representations, we detect anomalous time series using Clustering and the Wasserstein distance. Our results on the publicly available ECG5000 electrocardiogram dataset show the ability of the proposed approach to detect anomalous heartbeats in a fully unsupervised fashion, while providing structured and expressive data representations. Furthermore, our approach outperforms previous supervised and unsupervised methods on this dataset.
@article{Pereira2019BigComp, author = {Pereira, João and Silveira, Margarida}, citation_count_index = {1}, journal = {2019 IEEE International Conference on Big Data and Smart Computing (BigComp)}, title = {Learning Representations from Healthcare Time Series Data for Unsupervised Anomaly Detection}, year = {2019}, month = feb, publisher = {IEEE}, volume = {}, number = {}, pages = {1-7}, doi = {10.1109/BIGCOMP.2019.8679157}, url = {https://ieeexplore.ieee.org/document/8679157}, }
Unsupervised Anomaly Detection in Energy Time Series Data Using Variational Recurrent Autoencoders with Attention

João Pereira and Margarida Silveira

2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA), Dec 2018

Abs DOI Bib HTML PDF Slides Cited by 228

In the age of big data, time series are being generated in massive amounts. In the energy field, smart grids are enabling a unprecedented data acquisition with the integration of sensors and smart devices. In the context of renewable energies, there has been an increasing interest in solar photovoltaic energy generation. These installations are often integrated with smart sensors that measure the energy production. Such amount of data collected makes the quest for developing smart monitoring systems that can detect anomalous behaviour in these systems, trigger alerts and enable maintenance operations. In this paper, we propose a generic, unsupervised and scalable framework for anomaly detection in time series data, based on a variational recurrent autoencoder. Furthermore, we introduce attention in the model, by means of a variational self-attention mechanism (VSAM), to improve the performance of the encoding-decoding process. Afterwards, we perform anomaly detection based on the probabilistic reconstruction scores provided by our model. Our results on solar energy generation time series show the ability of the proposed approach to detect anomalous behaviour in time series data, while providing structured and expressive representations. Since it does not need labels to be trained, our methodology enables new applications for anomaly detection in energy time series data and beyond.
@article{Pereira2018ICMLA, author = {Pereira, João and Silveira, Margarida}, journal = {2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA)}, title = {Unsupervised Anomaly Detection in Energy Time Series Data Using Variational Recurrent Autoencoders with Attention}, year = {2018}, month = dec, publisher = {IEEE}, volume = {}, number = {}, pages = {1275-1282}, doi = {10.1109/ICMLA.2018.00207}, url = {https://ieeexplore.ieee.org/document/8614232}, }

Blog posts

All posts

2023	Fine-tune MPT-7B on Amazon SageMaker
2023	Fast and Scalable Hyperparameter Tuning and Cross-validation in AWS SageMaker

Career & education

adidas

Jheronimus Academy of Data Science

Eindhoven University of Technology

Instituto Superior Técnico, University of Lisbon

Highlights

Selected publications

Blog posts