Gabriel Gustavo Costanzo

We have developed an augmentation strategy that advances the field by integrating explainable constraints derived from ensemble model logic into the sample generation process. By relying on class-specific boundaries and employing a heuristic optimization routine grounded in genetic algorithms, our method ensures that all generated data points are semantically valid, diverse, and explicitly traceable to decision rules learned by the model. In doing so, it addresses a critical gap in the literature by offering a principled framework for explainable and constraint-compliant data augmentation.

Thesis supervisor: Sylvio Barbon Junior

• Hands-on experience with Deep Learning models.
• Hands-on experience with Python machine learning libraries, including scikit-learn and Pytorch.
• Capable of optimising and fine-tuning deployed predictive models for enhanced accuracy and stability.
• Managing assets (data, models, environments).
• Creating and debugging ML pipelines for feature engineering, training, and deployment.
• Hands-on experience with unsupervised learning techniques such as K-Means clustering.
• Python, SQL, Power M and DAX; unstructured data types and related file format (e.g. JSON).
• Azure Data Platform tools (SQL Server, Azure Databricks, Microsoft Fabric/Power BI).
• Knowledge of Data Lakes and Data Lakehouse architecture.
• Please, take a look at my project portfolio at the following link: https://github.com/GGustavoCostanzo?tab=repositories

• Project manager for Agroindustry Research Center.
• Key activities I have been involved with during this role:
• Project Identification, Formulation, and Design.
• Intellectual property advisor.
• Technology transfer advisor.
• Collaboration in the formulation of the project: "Technology and Innovation Management Master". ERASMUS-HORIZON2020. (585739-EPP-1-2017-1-AR-EPPKA2-CBHE-JP). November 2015 to February 2017.
• Website: https://inta.gob.ar/ciagroindustria

• ProBIO-Nativa project. Probiotics for low-scale, agro-ecologic production (reduction in the use of antibiotics, the transformation of agro-alimentary bi-products into high-quality animal feed).
• I managed to isolate lactic acid bacteria and yeasts with powerful antimicrobial activity, even higher than antibiotics commonly used to treat infections in livestock.
• Director: Dr. Guillermo Kociubinski
• Link: https://inta.gob.ar/documentos/produccion-porcina-inhibicion-de-enteropatogenos-antibiotico-multi-resistentes-porprobioticos-nativos-agricultura-familiar-vs-cria-intensiva

• BIOTECSUR platform (MERCOSUR-EU). Obtaining new antigens for the design of diagnostic tests and bovine vaccines.
• During this work, I published the following scientific article: Costanzo, G., Alvarado Pinedo, F., Mon M. L., Gioffré, A., Carrica Illia, M., Arese, A., Travería, G., y Romano, M. I. Accuracy assessment and screening of a dairy herd with paratuberculosis by three different ELISAs. Veterinary Microbiology. 2012. 156(1-2):183-8. DOI: 10.1016/j.vetmic.2011.10.029
• Director: Dra. María Isabel Romano
• Link: https://www.biotecsur.org/

• In the frame of the projects: 1- Role of NK KIR receptor and the polymorphism of cytokine promotors in the outcome of hepatitis B virus (HBV) and hepatitis C virus (HCV), and 2- Contribution of changes by intracellular signaling pathways to explain the immune response to antigens and alloantigens.
• During the course of this work I published the following scientific article: Paladino, N., Flores, A. C., Cherñavsky, A. C., Fainboim, H., Schroder, T., Cuarterolo, M., Goñi, J., Galoppo, C., Cañero-Velasco, M. C., Levi, D., Costanzo G. G., Arruvito, L., and Fainboim, L. The most severe forms of type I autoimmune hepatitis are associated with genetically determined levels of TGF-beta1. Clin Immunol. 2010. 134(3):305-12. DOI: 10.1016/j.clim.2009.11.004

• Undergraduate Thesis entitled "Combination of Adenovirus-shChk1 with Cytotoxic Stress Inducing Drugs to Kill Tumor Cells".
• I worked on evaluating a genetically modified adenovirus to attack tumor cells. I managed to demonstrate that the application of this biological drug proved to be very effective and showed that, in combination with very low levels of chemotherapeutic agents, tumor cells died while normal cells were preserved in good condition.
• Thesis supervisor: Dra. Vanesa Gottifredi. Head of the laboratory: Osvaldo Podhajcer.