Nicola Brazzale

Deep Learning Research Engineer

I multiply large matrices on GPUs for a living

Arnhem, the Netherlands

GitHubLinkedInX
NB

About

Ambitious Deep Learning Engineer with a track record of delivering end-to-end ML solutions from research to production. I thrive in mission-oriented, interdisciplinary teams and believe that any challenge can be overcome through dedication and a strong work ethic. I bring hands-on experience in model development, scalable pipeline engineering, and deploying systems in production and regulated environments. I value direct communication and collaboration as core strengths.

Work Experience

Thirona

2026 - Present

Senior Deep Learning Engineer

•Serve as technical point of contact for deep learning within the team, advising on design decisions and evaluating emerging technologies for adoption.

•Define the scope, content, and planning of algorithm releases, coordinating delivery from stakeholder requirements through to shipped features.

•Support and mentor Deep Learning Engineers in designing and developing new algorithm features.

•Reflect on the status of research projects in terms of technical performance and surface risks early.

•Drive systematic experimentation and ablation studies to validate architectural choices and model improvements.

Thirona

2023 - 2026

Deep Learning Engineer

•Designed and developed deep learning models (CNNs, UNets) for segmentation and classification tasks on 3D volumetric data.

•Owned the end-to-end ML lifecycle for allocated projects — from data collection and annotation to model training, evaluation, and initial integration into production systems.

•Built and maintained scalable pipelines covering preprocessing, training, inference, and evaluation.

•Optimised models for production deployment, reducing inference latency and improving robustness at scale.

Aalto University – ML4H Research Group

2021 - 2022

Research Assistant

•Conducted research on the comparison of ViTs and traditional CNNs for Chest X-Ray classification. Pre-trained a miniGPT model on a large corpus of radiology reports and fine-tuned it with visual features extracted from X-Rays to build a lightweight vision-language model.

•Evaluated multiple datasets and data augmentation strategies to analyse their effect on model efficiency and diagnostic performance.

Education

Aalto University

2020 - 2022
MSc in Machine Learning, Data Science and Artificial Intelligence
Major: Machine Learning: Advanced Probabilistic Methods, Bayesian Data Analysis, Gaussian Processes, Deep Learning, Kernel Methods, Computer Vision, and Data Mining. Bioinformatics minor: Computational Genomics, Machine Learning for bioinformatics, AI in health technologies, and Medical Image Analysis. Electives: Linear and non-linear optimisation.

University of Padua

2016 - 2019
BSc in Computer Engineering
Relevant courses: Algorithms and Data Structures, Database Management Systems, Optimisation, Artificial Intelligence, Embedded System Programming, and Computer Networks.

Technical Skills

Advanced Knowledge
Python
PyTorch
TensorFlow
OpenCV
Keras
Git
Docker
Good Knowledge
ITK
Jenkins
SQL
Basic Knowledge
Julia
Java
C++
R
Matlab

Projects

Artery-Vein Phenotyping – AVX

Contributed to the artery-vein segmentation module by researching practical improvements. Designed and implemented parts of the pipeline — including training, inference and evaluation — and explored different normalisation techniques, new metrics, and loss functions to accurately segment anatomical structures. Refined biomarker calculations to provide clients with precise and informative vascular measurements.

UNets
CNNs
TensorFlow
Docker

Lobes and Fissures Segmentation

Involved in the research and development of two segmentation models: one for pulmonary lobes and one for fissures. Achieved significant improvements in both inference speed and segmentation accuracy through architectural and preprocessing enhancements. Introduced evidential deep learning to produce calibrated uncertainty estimates; these confidence signals are used in production to flag low-quality outputs and inform downstream decision-making.

UNets
CNNs
TensorFlow
Docker
Evidential Deep Learning

Multi-modal Chest X-Ray Analysis

Improved downstream classification performance via self-supervised pre-training with BarlowTwins on large unlabeled X-Ray datasets. A GPT-based model was then fine-tuned with extracted visual features to generate radiology-style reports from X-rays.

Master Thesis
PyTorch
Lightning AI
Weights & Biases
HPC

Older University Projects

A collection of academic and side projects covering signal and image classification tasks — including ECG arrhythmia detection, lymphoma subtype classification, and EEG seizure classification. See GitHub for details.

PyTorch
TensorFlow
GANs
RNNs
CNNs
Matlab

Press Ctrl+J to open the command menu