VDA-SDLC: Validation-Driven Adaptive Agentic Software Development Life Cycle Framework for Autonomous Software Engineering
VDA-SDLC: Validation-Driven Adaptive Agentic Software Development Life Cycle Framework for Autonomous Software Engineering
Deepthi K1, Dr. Sunita Chalageri2
1 Department of Computer Science and Engineering & K S Institute of Technology
2 Department of Computer Science and Engineering & K S Institute of Technology
Abstract - This paper presents a Validation-Driven Adaptive Agentic SDLC Framework that automates multiple software engineering activities through a collaborative multi-agent architecture. The proposed framework integrates specialized agents responsible for requirement analysis, planning, code generation, validation, and packaging. Unlike conventional AI coding assistants, the framework employs a validation-driven feedback mechanism in which generated applications are automatically verified before deployment preparation. When inconsistencies or incomplete outputs are detected, corrective feedback is provided to the generation stage to improve software quality.
The framework utilizes OpenRouter-based Large Language Models including DeepSeek, Qwen, and GPT-OSS for frontend application generation. The architecture was implemented using FastAPI and evaluated through the generation of multiple frontend applications including calculators, weather dashboards, password generators, task management systems, and quiz applications. Experimental observations indicate that the proposed architecture successfully automates major frontend SDLC activities while improving application consistency and reducing manual development effort.
The proposed framework demonstrates the feasibility of autonomous software engineering using validation-driven agent collaboration and provides a foundation for future full-stack autonomous development environments.
Key Words: Agentic AI, Software Engineering, Multi-Agent Systems, SDLC Automation, Large Language Models, Autonomous Software Generation.