Jose Costa

LISN, CentraleSupelec, Paris-Saclay University

Anomaly detection, tabular data, self-supervised learning

CentraleSupelec, Paris-Saclay University

University of Sao Paulo

Interdisciplinary Laboratory of Numerical Sciences (LISN)

• Conducted a comprehensive analysis of anomaly detection models with thorough literature review and state-of-the-art data modeling techniques.
• Boosted classification and regression scores implementing self-supervised techniques in tabular data.

French Alternative Energies and Atomic Energy Commission (CEA)

• Reduced model size by 50% while maintaining accuracy by implementing a novel quantization scheme for Visual Transformers (ViT).
• Reduced computational complexity, making Visual Transformers more embeddable on resource-constrained devices.
• Presented results on attention-based network compression at the 2024 HiPEAC workshop.

Universite Paris-Cite

• Developed a novel graph-based method for embedding time series, improving anomaly detection accuracy by 22% on medical datasets.
• Identified anomalous behavior in multivariate medical time series by creating a latent space representation and constructing a graph to isolate outliers.
• Leveraged autoencoders and graph theory to design a novel anomaly detection method, reducing model training time.

Petrobras / University of Sao Paulo

• Improved mooring line failure detection accuracy by 18% by developing a machine learning pipeline using signal oscillation frequencies.
• Reduced detection time through signal processing techniques like Wavelets and STFT.
• Presented research at OMAE 2021 international conference.

University of Sao Paulo

Conferred on the program's top-ranking graduate for outstanding academic performance.

University of Sao Paulo

Highly competitive, university-wide scholarship awarded for exceptional academic excellence.

Oral · MICCAI 2026 @ MultiTab Workshop · Strasbourg, France

ASME OMAE 2024 · Singapore

Oral · 6th AccML Workshop, HiPEAC 2024 · Munich, Germany

Oral · Junior Multidisciplinary Congress, Universite Paris-Saclay · Paris, France

ASME OMAE 2021 · Online

LUSIS Chair / CentraleSupelec

Research supervisor for third-year engineering student projects on anomaly and fraud detection, from problem definition through to the final defense.

CentraleSupelec, Paris-Saclay University

Assistant in "Introduction to Python Programming" and "Algorithms".

University of Sao Paulo

Assistant in "Algorithms and Data Structures" and "Systems Programming".