The aim of this final project is to analyze the design, development, and managerial impact of Project P1, an advanced virtual assistant based on Large Language Models. The research frames the system not as a mere piece of software, but as an intangible productive asset within the Industry 4.0 paradigm. The methodology is structured across three levels: 1. Technological level: analysis of the modular architecture (Docker, microservices) and the data-processing pipeline (RAG, FAISS). 2. Process level: study of the software development life cycle (SDLC) and versioning strategies (GitFlow). 3. Managerial level: economic assessment through the profitability-diagram model to determine the break-even point (Q*). The results confirm the technical feasibility (Uptime > 99%) and the economic sustainability of the project. The analysis demonstrates that the Break-Even Point is reached within 4 months (April 2026) at 23,502 queries/year, with an estimated saving of over 600 man-hours per year and a positive ROI starting from the second year of operation.
L'obiettivo di questa prova finale è analizzare la progettazione, lo sviluppo e l'impatto gestionale del Progetto P1, un assistente virtuale avanzato basato su Large Language Models. La ricerca inquadra il sistema non come un mero software, ma come un asset produttivo immateriale all'interno del paradigma Industria 4.0. La metodologia adottata si articola su tre livelli: 1. Livello Tecnologico: analisi dell'architettura modulare (Docker, microservizi) e della pipeline di elaborazione dati (RAG, FAISS). 2. Livello Processuale: studio del ciclo di vita del software (SDLC) e delle strategie di versionamento (GitFlow). 3. Livello Gestionale: valutazione economica tramite il modello del diagramma di redditività per determinare il punto di pareggio (Q*). I risultati confermano la fattibilità tecnica (Uptime > 99%) e la sostenibilità economica del progetto. L'analisi dimostra il raggiungimento del Break-Even Point in 4 mesi (Aprile 2026) a 23.502 query/anno, con un risparmio stimato di oltre 600 ore/uomo annue e un ROI positivo a partire dal secondo anno di operatività.
Progetto P1: Ingegnerizzazione di un Impianto di Servizio Informativo basato su Large Language Models
LANDO, ANDREA
2025/2026
Abstract
The aim of this final project is to analyze the design, development, and managerial impact of Project P1, an advanced virtual assistant based on Large Language Models. The research frames the system not as a mere piece of software, but as an intangible productive asset within the Industry 4.0 paradigm. The methodology is structured across three levels: 1. Technological level: analysis of the modular architecture (Docker, microservices) and the data-processing pipeline (RAG, FAISS). 2. Process level: study of the software development life cycle (SDLC) and versioning strategies (GitFlow). 3. Managerial level: economic assessment through the profitability-diagram model to determine the break-even point (Q*). The results confirm the technical feasibility (Uptime > 99%) and the economic sustainability of the project. The analysis demonstrates that the Break-Even Point is reached within 4 months (April 2026) at 23,502 queries/year, with an estimated saving of over 600 man-hours per year and a positive ROI starting from the second year of operation.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_AL.pdf
accesso aperto
Dimensione
600.81 kB
Formato
Adobe PDF
|
600.81 kB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/115793