Bryan Guilcapi

Bryan Guilcapi

Auto-generated profile, based on publicly available information

Bryan Guilcapi

Institute of Applied Sciences and Intelligent Systems

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
13 March 2026 in Medinformatics
Verified
Journal Article
Open access
PDF file
BGBryan Guilcapi
Bryan Guilcapi
AMAlessia Milano
Alessia Milano
ADAmalia D'AVino
Amalia D'AVino
DSDomenico Sagnelli
Domenico Sagnelli
MRMassimo Rippa
Massimo Rippa
LPLucia Petti
Lucia Petti
Surface-enhanced Raman spectroscopy (SERS) combined with machine learning (ML) has emerged as a powerful strategy for intelligent and automated biosensing. In this study, we present an in-house fabricated and data-driven SERS platform for automated discrimination of multiple experimental states within a nanostructured biosensor system. The sensing architecture was developed using reproducible gold nanoparticle-coated silicon substrates (Au NPs), synthesized and assembled in-house, followed by functionalization with a Raman reporter molecule (4-mercaptobenzoic acid, 4-MBA) or three biologically relevant targets: Escherichia coli, OC43 coronavirus, and Shiga toxin 2. The classification task was formulated as a six-class problem corresponding to distinct experimental sensor states: bare silicon substrate, Au NP-coated plasmonic substrate, 4-MBA-plasmonic functionalized surface, and plasmonic substrates functionalized with each biological target. Raman spectral data were processed through an automated analytical pipeline and evaluated using principal component analysis, linear discriminant analysis, support vector machines, random forest, and a one-dimensional convolutional neural network (1D-CNN). Among the evaluated models, the 1D-CNN achieved superior performance, providing the highest classification accuracy and robust discrimination across all six experimental classes. The results demonstrate that deep learning applied directly to normalized spectral vectors enhances feature extraction and class separability compared to conventional approaches. This work highlights the potential of integrating in-house engineered SERS nanoplatforms with automated ML frameworks for comprehensive sensor-state discrimination and next-generation intelligent biosensor development. Received: 29 December 2025 | Revised: 20 February 2026 | Accepted: 28 February 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The clean data that support the findings of this study are openly available in public repositories: https://github.com/BryanGuilcapi/Machine-Learning-Enhanced-SERS-on-Silicon-Gold-Sensors-for-Ultra-High-Throughput-Pathogen-Detection. The original database is accessible by requesting the corresponding author. Author Contribution Statement Bryan Guilcapi: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision. Alessia Milano: Software, Validation, Investigation, Data curation, Writing - review & editing. Amalia D' Avino: Investigation, Writing - review & editing. Domenico Sagnelli: Validation, Investigation. Massimo Rippa: Investigation. Valentina Marchesano: Investigation. Ivan Salvatore Perrotta: Investigation. Rosa Luisa Ambrosio: Investigation. Giovanna Fusco: Investigation. Maurizio Brigotti: Investigation, Writing - review & editing. Stefano Morabito: Resources. Lucia Petti: Resources, Writing - review & editing, Supervision, Project administration, Funding acquisition.

20 April 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
FVFulvia Villani
Fulvia Villani
Gustavo ArdilaGustavo Ardila
Gustavo Ardila
FLFausta Loffredo
Fausta Loffredo
Anna De Girolamo Del MauroAnna De Girolamo Del Mauro
Anna De Girolamo Del Mauro
BGBryan Guilcapi
Bryan Guilcapi
GNGiuseppe Nenna
Giuseppe Nenna
Photo-actuators coupled with mechanical energy harvesting systems feature a promising approach to solar energy conversion in the broad field of innovative energy harvesting technologies. In this study, we present a novel device capable of generating a 0.2 V peak-to-peak signal by sustaining the continuous oscillation of photomobile polymer (PMP) films in the range of 5-8 Hz when exposed to continuous steady illumination by effectively transferring mechanical motion to a piezoelectric transducer (PZL). An optomechanical energy-harvesting prototype based on PMPs was fabricated and experimentally characterized utilizing Xenon arc-lamp that closely mimics natural sunlight electromagnetic radiation. The analyzed harvesting system exhibits a distinctive dual-stage energy conversion process, from light to mechanical and subsequently from mechanical to electrical energy, highlighting its potential for diverse applications in next-generation solar energy systems.

11 September 2026 in Machine Learning: Science and Technology
Verified
Journal Article
Open access
Domenico SagnelliDomenico Sagnelli
Domenico Sagnelli
BGBryan Guilcapi
Bryan Guilcapi
AMAlessia Milano
Alessia Milano
ADAmalia D’AVino
Amalia D’AVino
VMValentina Marchesano
Valentina Marchesano
LPLucia Petti
Lucia Petti
Although autonomous, large language model-driven systems show immense potential for orchestrating complex scientific experiments, their efficacy is constrained by two fundamental bottlenecks: context dilution, where strategic reasoning degrades as experimental history accumulates, and inter-campaign amnesia, which forces systems into computationally expensive tabula rasa explorations upon encountering novel domains. To overcome these limitations, we introduce the Multi-Objective State-Action Network (MOSAN), an autonomous cognitive information fusion framework designed for robust cross-campaign generalization and sensor shift adaptation. MOSAN achieves effective information fusion by coupling a self-evolving semantic memory with an Organic Strategic Graph Memory (OSGM), strictly orchestrated through a novel Single Ledger Architecture that isolates cognitive phases and prevents context saturation. A core mechanism of the OSGM is Cross-Domain Strategic Seeding (‘Ghost Node’ injection), a transient cold-start fusion strategy. Upon encountering novel datasets, the OSGM temporarily injects topological priors from similar past domains; these ephemeral nodes guide initial architectural deductions and are subsequently purged, allowing the system to disseminate structural knowledge across disciplines while decreasing the risk of cross-contamination. The fusion framework was validated on the challenging task of bacterial classification via Surface-Enhanced Raman Spectroscopy subject to sensor aging shifts, autonomously discovering highly accurate multi-topological architectures. Crucially, when subjected to an out-of-distribution 1D sequential modality (the ECG5000 cardiovascular dataset), the agent performed zero-shot architectural meta-learning, bypassing brute-force search to rapidly achieve 0.9930 accuracy. Finally, the framework was successfully deployed using an open-weight model (Mistral 24B) at zero API inference cost. By continuously fusing multi-epoch empirical evidence, bridging representational topologies, and actively overcoming sensor drift, MOSAN establishes a scalable and rigorous paradigm for autonomous biophysical discovery.