Berardino Barile Ph.D.

Current Roles and Affiliations

Lead AI Engineer at Fastweb S.p.A. (Rome, Italy):

Berardino Barile is a Lead AI Engineer at Fastweb S.p.A. (Rome, Italy), where he leads the design, development, and deployment of Large Language Model (LLM) and agentic automation systems. He drives initiatives that take AI solutions from research prototypes to production-grade services, automating complex, knowledge-intensive workflows end-to-end.

His current work centers on multimodal Vision-Language Model (VLM) Retrieval-Augmented Generation (RAG) for document intelligence, steering and fine-tuning of LLMs (activation steering, instruction tuning, and parameter-efficient methods such as LoRA/PEFT), and building agentic, voice-driven assistants that orchestrate tool use and function calling to autonomously execute operational tasks. This bridges cutting-edge Generative AI research with real-world enterprise applications.

Previous Professional Experience

Berardino Barile has a rich background spanning both academic and industry positions. Prior to his current role, he held research and data science positions across Canada and Italy, working in sectors from AI research to public policy and insurance. Notable past positions include:

• Research Scientist at McGill & Concordia University – Mila AI Institute, Canada (2023–2025): Led AI research projects in computer vision within the Probabilistic Vision Group (McGill), the IMPACT Lab (Concordia), and Mila – Quebec's AI Institute (Yoshua Bengio), under the supervision of Prof. Tal Arbel. He leveraged generative AI (LLMs) and Causal ML for zero-shot prediction and causal reasoning in longitudinal disease evolution, and served as a teaching assistant at McGill in probability theory, Bayesian inference, and deep generative models. He presented research at top-tier conferences (MICCAI, ICCV, NeurIPS, CVPR).
• Senior Data Scientist at Verti S.p.A. (Mapfre Group), Italy (2022–2023): Led machine learning initiatives in an insurance context, such as optimizing pricing models and deploying automated ML pipelines on cloud platforms (AWS SageMaker) . He also collaborated with Sapienza University of Rome and the Italian workers’ compensation authority (INAIL) on modeling workplace accidents and business failure using causal inference and predictive modeling .
• Big Data Scientist at Isiway S.r.l., Italy (2017–2019): Implemented data-driven solutions for large-scale consulting projects, focusing on big data processing, parallelizing machine learning models on high-dimensional datasets, and advanced data visualization. During this time, he enjoyed contributing to company tech blogs – for example, he authored an article on deep NLP techniques for hate speech detection in social media.
• Data Analyst / Research Scientist at Invitalia (National Agency for Investment and Economic Development), Italy (2014–2017): Provided statistical analysis and policy evaluation research to support public management decisions . He developed econometric models and counterfactual analyses to assess the impact of public investment incentives and interventions . In this role, he co-authored the Annual Statistical Report on the Incentives Issued to the Italian Economic System for the Italian Ministry of Economic Development, and contributed to studies on entrepreneurship and finance (see Publications). His work at Invitalia led to presentations at international conferences and workshops on policy evaluation.
• Data Analyst (Intern) at Johnson & Johnson Medical (MedTech), Italy (2013–2014): Worked on healthcare data analytics developping statistical models and mathematical optimization as first post-graduate experience following his M.Sc.

These roles illustrate Barile’s blend of public-sector experience and private-sector data science. Across these positions, he honed skills in statistical analysis, machine learning deployment, big data tools, and interdisciplinary collaboration.

Education and Academic Background

Berardino Barile has a strong academic foundation in statistics and machine learning, including a double doctoral degree. His educational qualifications are as follows:

• Ph.D. in Engineering Science (Machine Learning), KU Leuven (Belgium): Awarded 2022. This was Barile’s second doctoral degree, earned as part of a joint program. His Ph.D. research at KU Leuven was conducted in parallel with his work in France (as a co-tutelle program), under a prestigious Marie Skłodowska-Curie fellowship.
• Ph.D. in Biomedical Engineering and Imaging Science, Université Claude Bernard Lyon 1 (France): Awarded 2022. This was his first Ph.D. (completed jointly with KU Leuven). His doctoral thesis, titled “Machine Learning Methods for Multiple Sclerosis Classification and Prediction using MRI Brain Connectivity,” focused on applying advanced machine learning to neuroimaging data for multiple sclerosis (MS) research . During this project, he was based at the CREATIS research center in Lyon and worked closely with medical experts. He received a Marie Curie Fellowship to support this dual-PhD endeavor .
• M.Sc. in Statistics, Sapienza University of Rome (Italy): Completed 2013 with highest honors (110/110 cum laude). His master’s thesis dealt with international banking systems and macroeconomic analysis (using Structural-VAR models), reflecting his early interest in econometrics and finance.
• B.Sc. in Statistics, Sapienza University of Rome: Completed 2011 with highest honors (110/110 cum laude). His bachelor’s thesis examined Purchasing Power Parity (PPP) and equilibrium exchange rates in financial markets , indicating a foundation in economic statistics.

In addition to formal degrees, Barile has pursued continuous learning through workshops and summer schools (especially in Machine Learning and Artificial Intelligence, advanced statistical modelling, econometrics, and public policy evaluation) during his early career. His multilingual ability (Italian native, English advanced, French advanced, Spanish beginner) and international study/work experience in Italy, France, Belgium, and Canada highlight a global academic perspective.

Technical Skills and Areas of Expertise

Combining a strong foundation in statistics and computer science with hands-on production AI engineering, Barile’s skill set spans a wide range of technical areas, with a current focus on Generative AI, large language models, and multimodal reasoning:

• Generative AI & Large Language Models (LLMs): Designs, builds, and deploys production-grade LLM systems end-to-end. He works on LLM steering and fine-tuning (activation steering, instruction tuning, and parameter-efficient methods such as LoRA/PEFT) to align open- and closed-source models with domain-specific requirements, improving reliability, safety, and downstream task performance. He is experienced with context engineering, harness engineering, and loop engineering for agentic LLM systems, along with model evaluation and grounding techniques to reduce hallucinations and boost factual accuracy.
• Multimodal Reasoning & Vision-Language Models (VLMs): Develops and optimizes multimodal Retrieval-Augmented Generation (RAG) pipelines that jointly reason over visual and textual content for document intelligence, grounding model responses in enterprise knowledge bases. He combines vision-language models with retrieval and embedding techniques to enable cross-modal understanding of images, documents, and text.
• Agentic AI & Automation: Engineers multi-agent and agentic systems that automate complex, knowledge-intensive workflows end-to-end, orchestrating tool use, function calling, and planning. This includes building real-time, voice-driven assistants (integrating speech-to-text / text-to-speech) that autonomously execute operational tasks, taking solutions from research prototypes to deployed services.
• Machine Learning & Deep Learning: Expertise in statistical learning, data mining, and deep learning. He has worked with supervised and unsupervised learning, ensemble methods, neural networks (including generative adversarial networks, transformers, and reinforcement learning), and causal inference techniques, spanning computer vision, NLP, and time-series applications.
• Causal Inference & Uplift Modeling: Barile has applied causal inference methods to both medical treatment effect estimation and policy impact evaluation. He works on Individual Treatment Effect (ITE) and Conditional Average Treatment Effect (CATE) estimation for personalized medicine. In the policy domain, he employs uplift modeling (to capture heterogeneous treatment effects) in evaluating interventions (as seen in his 2024 study on Causal Machine Learning for safety policies).
• Computer Vision & Medical Image Analysis: Through his Ph.D. and research career, he developed deep specialization in computer vision and biomedical imaging, especially MRI-based brain connectomics and image segmentation for neurological diseases (multiple sclerosis). He is familiar with advanced image processing, tensor analysis, generative models for imaging, and graph-based learning on brain networks.
• Programming, MLOps & Tools: Proficient in Python and MATLAB (advanced level), with experience in R, Stata, SQL, and LaTeX. He is skilled in modern ML frameworks (PyTorch, TensorFlow) and in deploying models to production using cloud and MLOps tooling: AWS SageMaker, Docker, Linux/Bash, and CI/CD for scalable training, serving, and workflow automation.
• Domains of Expertise: Barile’s unique combination of skills allows him to work at the intersection of enterprise AI (LLMs, agentic automation, document intelligence), healthcare (medical imaging, personalized medicine), public policy and economics (industrial policy, startup financing), and core AI research (generative models, graph neural networks). This interdisciplinary range is evident in his publications and projects, where he collaborates with engineers, neurologists, economists, and computer scientists alike.

Research Projects and Contributions

Berardino Barile has been involved in numerous research projects across different fields. Below is an overview of key projects, with their objectives, collaborators, and outcomes:

1. Machine Learning for Medical Imaging in Multiple Sclerosis (Ph.D. Projects, 2019–2022)

During his double Ph.D., Barile’s main focus was on leveraging machine learning to improve the understanding and prediction of multiple sclerosis using MRI data. Working with advisors Prof. Sabine Van Huffel (KU Leuven) and Dr. Dominique Sappey-Marinier (Lyon 1), and in collaboration with researchers like Aldo Marzullo, Claudio Stamile, and Pooya Ashtari, he contributed to several sub-projects:

• GAN-based Data Augmentation for Brain Connectomics: He authored a project to address limited sample sizes in MS studies by generating synthetic brain connectivity data. The team developed a Generative Adversarial Network (GAN) framework to create realistic synthetic brain structural networks from MRI, augmenting the connectome data of MS patients. The study demonstrated that augmenting training datasets with GAN-generated samples significantly improved the accuracy of classifying MS disease subtypes – for example, the classification F1-score increased from 66% to 81% when using augmented data versus original data . This work, published in Computer Methods and Programs in Biomedicine (2021), validated GANs as a valuable tool for biomedical data augmentation.
• Ensemble Learning for Disability Prediction: Barile was first author of a study that proposed an ensemble of gradient boosting models (GBM, XGBoost, CatBoost, LightGBM) to estimate MS patient disability as measured by the Expanded Disability Status Scale (EDSS) . By using brain structural connectivity metrics derived from diffusion MRI, the model could predict a patient’s disability score. The ensemble was augmented with an interpretable component (a logistic regression) to identify which brain connections most influenced the disability prediction . The approach achieved high accuracy (a low RMSE of ~0.92 in EDSS prediction) and highlighted critical white-matter network links associated with higher disability . This research, published in Brain Connectivity (2022), illustrated how combining black-box models with interpretable models can yield both performance and insight in medical AI.
• Unsupervised Classification of MS Clinical Forms: In another project, he explored unsupervised learning to differentiate MS clinical types. Barile and colleagues applied Non-Negative Tensor Factorization (NTF) to longitudinal MRI connectivity data . The idea was to decompose 3D connectivity matrices over time into a set of latent factors capturing patterns of brain changes. These latent factors were then used to cluster patients into their clinical categories (Relapsing-Remitting, Secondary Progressive, etc.) without manual labels. The method successfully identified meaningful patterns of disease progression and the brain regions most affected by long-term changes. This work was presented at the ICPR 2022 conference (International Conference on Pattern Recognition 2022), demonstrating the potential of tensor factorization for neurological data analysis.
• Kernel-Based Multi-view Learning: Extending the above, he also investigated kernel methods combined with tensor decomposition to improve MS subtype classification. In one study (published in the ESANN 2022 conference proceedings), Barile et al. introduced a kernelized multilinear SVD approach to integrate multi-modal MRI data for MS profile classification . By using kernel functions, they generated feature embeddings that enhanced classification performance while using only standard anatomical MRI (T1-weighted images). This offered a simple but effective way to differentiate MS subgroups without requiring more expensive or complex imaging modalities.
• Deep Learning for Lesion Segmentation: Barile contributed to developing deep learning models for automatic MS lesion segmentation in MRI scans, an important task for tracking disease progression. He co-authored the design of Pre-U-Net, a 3D convolutional neural network with pre-activation residual blocks, aimed at segmenting new lesions in longitudinal FLAIR MRI sequences. By combining different data augmentation methods and a deep supervision training strategy, Pre-U-Net improved detection of small, newly formed lesions. In evaluations on the public MSSEG-2 challenge dataset, the model outperformed conventional U-Net variants, achieving an F1 score of ~48% for new lesion detection (and significantly higher sensitivity than baseline models). These results were published in Frontiers in Neuroscience (2023) and presented at the ISMRM conference. This work helps neurologists by providing an automated tool to identify disease activity over time.
• Imaging Biomarkers and Clinical Correlates: Barile has also been involved in interdisciplinary research examining quantitative MRI biomarkers. For instance, he co-authored a study on the T1/T2 ratio as a marker of brain tissue integrity in MS (with collaborators in Lyon). This study found that the T1/T2 intensity ratio in normal-appearing white matter can detect subtle microstructural damage in MS patients nearly as effectively as advanced diffusion tensor imaging measures. Such findings, published in 2022, highlight Barile’s contributions to biomedical research that directly ties imaging data to clinical insights.

Overall, through these projects, Barile significantly advanced the application of AI in multiple sclerosis research. He co-authored at least half a dozen publications in this domain between 2020 and 2023, ranging from journals like Scientific Reports and Frontiers to international conference proceedings. His work in this area is characterized by a combination of technical innovation (GANs, tensor factorization, deep CNNs) and practical medical relevance (improving disease monitoring and patient stratification).

2. Causal Machine Learning for Personalized Recommendations and Treatment (Research Scientist, 2023–2025)

• Individual Treatment Effect (ITE) Estimation: Barile explored methods to estimate how specific actions or “treatments” affect individual outcomes – for example, how a particular medical intervention might benefit one patient vs. another. This involved causal modeling techniques such as uplift modeling, counterfactual prediction, and causal forests. His background in both healthcare and economics positioned him to tackle ITE estimation using diverse data types (imaging, tabular clinical data, etc.).
• Personalized Recommendation Systems: Another application area was using generative models and causal inference for recommendations. For instance, Barile worked on recommendation algorithms that understand why they make a certain suggestion, not just what to suggest. By infusing causal reasoning, such systems can avoid spurious correlations and provide recommendations that would most benefit a specific user. He investigated generative models for personalized recommendation and how to evaluate recommendations as individualized “treatments” with measurable effects – a relatively new synergy in the field, combining recommender systems with causal effect estimation to improve personalization.
• Causal Reinforcement Learning: Barile explored the “integration of Causal Inference and Reinforcement Learning” as a promising synergy. In practical terms, this meant designing RL agents that understand causal structures (e.g. an agent that can distinguish cause and effect in an environment, leading to better generalization or safer decisions), in collaboration with Mila researchers in RL and causality. It aligned with his broader goal of AI that can reason about interventions and outcomes.

These projects, carried out during his time as a Research Scientist at Mila / McGill, reflect Barile’s commitment to advancing AI methodologies (causal ML, RL) with high-impact applications (recommenders, healthcare decision support, finance). Building on this foundation, his current work as Lead AI Engineer at Fastweb extends these methods into production-grade LLM and agentic systems.

3. AI for Public Policy and Economic Analysis (2018–2024)

Beyond pure technical AI research, Berardino Barile has actively worked on projects at the intersection of data science and public policy/economics:

• Startup Financing and Survival Analysis: Building on his Invitalia experience, Barile co-authored a study examining how early access to bank credit affects the long-term survival of startups. Using a longitudinal dataset of over 49,000 Italian startups founded in 2003–2005, the study investigated whether receiving bank loans in the startup’s early stage had a causal effect on its probability of default years later. The analysis employed instrumental variables (2SLS regression) to account for endogeneity. The key finding was that greater initial bank debt had a significant negative impact on firm survival, suggesting that heavy reliance on loans may impair a startup’s resilience. This research, published in Contemporary Economic Policy (2021), was done in collaboration with economists Prof. Angelo Castaldo and Giuliana De Luca. It provided evidence to inform how public loan guarantee programs should be structured to truly benefit new enterprises.
• Occupational Safety Policies Impact Evaluation: More recently, Barile is the lead author of an interdisciplinary project evaluating occupational safety and health (OSH) investment policies and their impact on firms’ outcomes. This project was a collaboration with Sapienza University (Dept. of Economic Studies) and resulted in a 2024 paper in Scientific Reports (Nature). The study carried out a causal impact evaluation of a government aid scheme in Italy that subsidized SMEs to improve workplace safety, assessing whether this intervention affected the firms’ survival rates. Barile and co-authors compared thirteen ML modeling approaches (including traditional econometric models and modern machine learning models) to predict firm default, and used uplift modeling to estimate the Individual Treatment Effect of the safety investment on each firm. The results showed measurable heterogeneity: for some firms, the policy significantly reduced default risk, while for others the effect was negligible – underlining the importance of targeted policy design. The best-performing model was a LightGBM (gradient boosting machine), and it achieved high uplift evaluation scores (AUUC and Qini coefficients) in identifying which firms benefited most. This work not only provided insights to policymakers about the efficacy of OSH incentives, but also showcased how causal ML techniques can augment policy analysis traditionally done with econometrics. (Notably, Barile shared the code publicly, as announced in his LinkedIn post about this paper, reflecting his commitment to open science.)
• Other Economics and Policy Studies: Barile has continued to engage in research on topics like workplace accidents and firm performance. He has co-authored (with Prof. Angelo Castaldo et al.) a paper on causal analysis of workplace accident prevention and firm default, and has contributed to studies on regional differences in workplace accident rates in Europe. He also co-wrote policy reports for the European Commitee (UE), such as the annual incentive report mentioned earlier, and has a book chapter (2023) on startup financing policies in Italy (as part of an edited volume on the Italian economy). These works further underscore his role in applying advanced analytics to socioeconomic data.

In summary, Barile’s research projects range from developing novel AI algorithms (for vision and causal learning) to applying data science for social good (health and economic policy). His ability to collaborate across disciplines is a recurring theme: whether it’s working with neurologists on medical AI or with economists on causal inference, his contributions have been significant and well-documented in publications.

Publications and Conference Contributions

Berardino Barile has an extensive list of publications in journals, conferences, and other outlets. Below is a selection of his notable academic publications and conference papers, along with brief summaries and context for each:

• Causal Impact of Safety Policies on Firms (Scientific Reports, 2024): Barile B., Marco Forti, Alessia Marrocco, Angelo Castaldo. “Causal impact evaluation of occupational safety policies on firms’ default using machine learning uplift modelling.” Published in Scientific Reports 14, 10380 (2024). This article presents a machine learning-based analysis of a workplace safety subsidy program, comparing treated vs. control firms. It demonstrated that machine learning models (especially LightGBM) can estimate policy effects at the firm level, uncovering that safety investments causally improved survival rates for certain SMEs. (This interdisciplinary work ties AI with economics and was featured in Barile’s public posts.)
• GANs for Brain Connectivity Data Augmentation (Comp. Methods Programs Biomed., 2021): Barile B., Aldo Marzullo, Claudio Stamile, Françoise Durand-Dubief, Dominique Sappey-Marinier. “Data augmentation using generative adversarial neural networks on brain structural connectivity in multiple sclerosis.” Comput. Methods Programs Biomed. 206:106113 (2021). In this paper, Barile et al. developed a GAN to generate synthetic brain network matrices. The augmented data helped improve machine learning performance in classifying MS patient groups. Notably, adding GAN-generated samples boosted a classifier’s F1-score from 66% to 81%, highlighting the benefit of synthetic data in biomedical applications.
• Ensemble Learning for MS Disability (Brain Connectivity, 2022): Barile B., et al. “Ensemble Learning for Multiple Sclerosis Disability Estimation Using Brain Structural Connectivity.” Brain Connectivity 12(5): 359–369 (2022). This work introduced a stacked ensemble of four boosting models to predict the disability level (EDSS) of MS patients from diffusion MRI connectome features. It achieved a high accuracy (RMSE ~0.92) in estimating disability and incorporated an interpretable model to identify key brain connections associated with higher disability scores. The study provided a more transparent AI approach in clinical decision support.
• Unsupervised MS Subtype Classification (ICPR 2022): Barile B., et al. “Tensor Factorization of Brain Structural Graph for Unsupervised Classification in Multiple Sclerosis.” In Proc. 25th Intl. Conf. on Pattern Recognition (ICPR), 2022. This conference paper proposed an unsupervised learning pipeline using non-negative tensor factorization to detect longitudinal brain connectivity changes and cluster MS patients by disease course. It was one of the first applications of tensor factorization for MS profiling, offering a new way to discover data-driven patient subgroups.
• MS Lesion Segmentation with Pre-U-Net (Frontiers in Neuro., 2023): Ashtari P., Barile B. (co-first author), et al. “New multiple sclerosis lesion segmentation and detection using pre-activation U-Net.” Frontiers in Neuroscience 17: 818675 (2023). This journal article presents the Pre-U-Net model for automated segmentation of new MS lesions in 3D MRI. The model outperformed baseline U-Net variants in the MSSEG-2 challenge, achieving about 40% Dice score for new lesion segmentation and significantly improving detection sensitivity. This helps in monitoring MS by reducing the burden of manual lesion annotation.
• MS Clinical Profiles Classification (Frontiers in AI, 2023): Barile B., et al. “Classification of multiple sclerosis clinical profiles using machine learning and grey matter connectome.” (Frontiers in Artificial Intelligence, 2023). In this study, Barile and colleagues built an automated pipeline to classify patients into clinical MS types (CIS, RRMS, SPMS, PPMS) using features derived from cortical thickness networks (grey matter connectome). By extracting graph metrics from MRI data and using a voting ensemble of simple ML classifiers, they achieved up to 86% F1-score in distinguishing certain MS forms. This result is notable because it used conventional MRI (no advanced sequences) and still provided high accuracy, suggesting a cost-effective tool for clinical profiling.
• Kernel ML for MS Profiles (ESANN 2022): Barile B., Pooya Ashtari, Françoise Durand-Dubief, Dominique Sappey-Marinier. “A Kernel Based Blind Source Separation Approach for Classification of MS Clinical Profiles.” In Proc. 30th European Symp. on Artificial Neural Networks (ESANN), Bruges, 2022. This conference paper combined kernel methods with multilinear SVD (a form of blind source separation) to classify MS patient profiles using only T1-weighted MRI. It showed that applying a kernel trick to multi-view neuroimaging data yields a meaningful feature embedding and good classification performance, even with simple linear classifiers. It reinforced the idea that sophisticated deep models are not always necessary if data is cleverly preprocessed.
• MRI Biomarker Research (NeuroImage: Clin., 2022): Hamad A., A. Sormani, C. Collorone, B. Parisi, B. Barile, et al. “T1/T2 ratio: a quantitative sensitive marker of brain tissue integrity in multiple sclerosis.” NeuroImage: Clinical 35: 103069 (2022). (Barile as co-author) This study validated the T1/T2 MRI signal ratio against diffusion tensor imaging metrics in MS patients, finding the ratio to be highly sensitive to white matter damage. Barile’s contribution here was related to data analysis, reflecting his collaboration with neuroimaging experts to identify new clinical biomarkers.
• Startup Bank Loans & Default (Contemporary Economic Policy, 2021): Castaldo A., De Luca G., Barile B. “Does initial access to bank loans predict start-ups’ future default probability? Evidence from Italy.” Contemporary Economic Policy 39(1): 88–106 (2021). In this economics paper, the authors analyzed whether getting bank loans in a startup’s early phase has a lasting effect on its survival. The study concluded that startups with easier early access to bank credit actually showed a higher likelihood of default later on, after controlling for various factors. This counter-intuitive finding (early debt can be a liability) provided important insights for entrepreneurship policy. Barile co-authored the statistical analysis and methodology, leveraging his statistics background in an economics context.

(The above list is not exhaustive; Barile has additional conference abstracts and papers, including policy research papers under review. His Google Scholar profile​ lists further details of his publications and citations.)

Technical and Personal Highlights

• Technical Strengths: Through these publications and projects, Barile has demonstrated strong capabilities in statistical analysis, machine learning engineering, and scientific computing. For example, implementing a full stack of machine learning pipelines at Verti indicates his ability to deploy models in production environments, while his academic work shows strength in experimental design and data-driven discovery. He is adept at programming (Python, MATLAB, R) and handling large datasets, and comfortable working in Linux/Bash and cloud setups. His dual expertise in classical statistics (e.g. regression, hypothesis testing, causal inference) and modern AI (deep learning, GANs, etc.) is a distinctive asset.
• Collaboration and Leadership: Barile often takes lead author roles (e.g., first author on the Scientific Reports 2024 paper and several MS studies), indicating his leadership in research projects. At the same time, his collaborations span multiple countries and fields – he has worked with medical doctors, statisticians, and economists. This interdisciplinary teamwork ability was nurtured by being part of labs like CREATIS in Lyon (for medical imaging) and by engaging with Sapienza’s economic studies department. He is also involved in mentorship and group activities; as a Research Scientist he co-supervised junior students in the Probabilistic Vision Group and contributed to Mila’s research community events.
• Open Science and Community Engagement: Barile appears to support open science principles. He has made code from his research available (e.g., releasing code for the uplift modeling study on GitHub). He also shares insights through online articles – for instance, his Medium blog post on deep NLP for hate speech aimed to explain AI methods to a broader audience. He also created a personal blog post called BNormal where he shares intriguing insights on subjects like Statistics and Computer Science. This suggests he values science communication and practical impact. On social media (e.g., LinkedIn), he discusses the implications of his research (such as rethinking the “sticks or carrots” approach in safety regulation).
• Personal Interests: While Barile’s public presence is largely professional, a few personal details can be inferred from his background. Having lived in Italy, France, Belgium, and now Canada, he enjoys multicultural experiences and speaks multiple languages fluently (Italian, English, French). His passion for “real-world problems” suggests an interest in applied challenges – he is motivated by using AI for social and healthcare good. His inclination to write explanatory articles hints at an interest in teaching or outreach. Given his statistical expertise, he follows developments in technology, economics, and finance.