Francesco Alzate
Automating the Software Development Lifecycle with Cooperative AI Agents and Human-in-the-Loop Supervision.
Rel. Paolo Garza. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering), 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (8MB) | Preview |
Abstract
The growing capability of Large Language Models to generate, analyse, and critique source code has opened new possibilities for automating software development workflows. However, the gap between generating isolated code snippets and reliably delivering complete, tested, and documented software artefacts remains substantial. This thesis addresses that gap through the design, implementation, and evaluation of MAST (Multi-Agent Software Team), an agentic framework that orchestrates a cooperative pipeline of specialised AI agents to automate the full Software Development Lifecycle. The pipeline comprises seven agents — Clarification, Specification, Implementation, Test, Critic, Documentation, and GitLab Push — each responsible for a well-defined stage of the development process.
All inter-agent communication is mediated exclusively through structured JSON artefacts stored on Amazon S3, a design that decouples agents from one another, enables independent testability, and provides a complete audit trail of every pipeline run
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
