SDLC Life Cycle

End to End Project Orchestration using Amazon Nova aka SDLC Life Cycle

Comment

Inspiration

Modern software development often requires weeks of planning, architecture design, coding, testing, and reviews before a working system is produced.

Many teams struggle with:

  • Translating raw ideas into structured requirements
  • Designing scalable system architectures
  • Maintaining consistent documentation
  • Ensuring quality and security from the start

We asked a simple question:

What if AI could act like a complete engineering team and execute the entire Software Development Lifecycle automatically?

Inspired by the power of Amazon Nova models and Agentic AI, we built Nova SDLC Architect — a platform that transforms an idea, voice input, or whiteboard sketch into a production-ready software project within minutes.

What it does

Nova SDLC Architect converts a plain-language requirement, voice description, or image-based design into a complete software project by orchestrating eight specialized Amazon Nova AI agents, each responsible for a stage of the Software Development Lifecycle.

The system automatically generates:

  • Structured requirements and user stories
  • System architecture and design decisions
  • Database schema and ER diagrams
  • API specifications
  • Complete application source code
  • UI/UX screen designs
  • Code review and security analysis
  • Automated test suites

Key Features

  • Live Token Streaming
    Watch Amazon Nova generate responses in real-time using streaming tokens.

  • Voice Input
    Capture requirements naturally using browser speech recognition.

  • Multimodal Image Analysis
    Upload whiteboard sketches, diagrams, or screenshots to automatically extract requirements.

  • Agent Handoff Visualization
    Visual feedback showing which AI agent is currently executing.

  • ZIP Download
    Download the full generated project ready to run locally.

This dramatically reduces the time required to move from idea → working system.

How we built it

The platform uses Amazon Bedrock Nova models orchestrated through a Python backend pipeline.

Technology Stack

Frontend

  • Vanilla JavaScript Single Page Application
  • Custom dark-themed UI
  • Web Speech API for voice input
  • WebSockets for real-time streaming updates

Backend

  • FastAPI
  • Python 3.10+
  • Uvicorn server
  • Pipeline session management

AI Layer

  • Amazon Bedrock
  • Nova Pro v1
  • Nova Lite v1
  • boto3 SDK

AI Agent Pipeline

The system coordinates eight specialized AI agents:

Agent Nova Model Output
Requirements Analyst Nova Pro v1 User stories, DoR/DoD, stakeholder mapping
Solution Architect Nova Pro v1 Architecture diagrams, ADRs, SLOs
Data Architect Nova Pro v1 Database schema, ERD, migrations
API Designer Nova Pro v1 OpenAPI specifications
Senior Developer Nova Pro v1 Full source code
UI/UX Designer Nova Lite v1 UI mockups and design systems
Code Reviewer Nova Lite v1 Security findings and complexity analysis
QA Automation Nova Pro v1 Unit, integration, e2e, and performance tests

Each agent receives the structured output from the previous stage, enabling a complete Agentic SDLC pipeline.

Amazon Nova Integration

API Used Purpose
bedrock-runtime.converse_stream() Real-time token streaming
bedrock-runtime.converse() Image analysis and fallback calls
Nova Pro Multimodal Image-based requirements extraction
Model: amazon.nova-pro-v1:0 Core development tasks
Model: amazon.nova-lite-v1:0 UI design and code review

All generated artifacts are fully produced by LLM agents.

Challenges we ran into

Agent Context Handoff

Passing structured context between eight agents without losing information required careful prompt engineering.

Real-Time Streaming

Implementing smooth real-time token streaming using WebSockets while maintaining UI responsiveness.

Multimodal Parsing

Extracting meaningful requirements from whiteboard images and diagrams was complex due to inconsistent drawing styles.

Maintaining Consistency

Ensuring generated architecture, APIs, database, and code stayed aligned across agents required validation steps.

Demo Mode Support

We implemented a fallback demo mode so the platform works even without AWS credentials.

Accomplishments that we're proud of

  • Built a fully functional 8-agent autonomous SDLC pipeline
  • Integrated Amazon Nova streaming APIs
  • Implemented multimodal requirement extraction
  • Added voice-driven requirement input
  • Generated complete runnable software projects
  • Built a real-time AI workflow visualization system
  • Demonstrated Agentic AI orchestration across the entire SDLC

This project shows how AI can move beyond code generation to full lifecycle engineering.

What we learned

Specialized AI Agents Work Better

Breaking SDLC tasks into specialized agents significantly improved results.

Transparency Builds Trust

Showing token streaming allows users to understand AI reasoning.

Multimodal Input Is Powerful

Developers frequently start with diagrams rather than written requirements.

Prompt Engineering Is Critical

Structured prompts ensured consistency across architecture, database, API, and code generation stages.

What's next for SDLC Life Cycle

We plan to evolve Nova SDLC Architect into a fully autonomous AI engineering platform.

CI/CD Integration

  • Automatic GitHub repository creation
  • AI-generated CI/CD pipelines
  • Cloud deployment automation

Cloud Architecture Generation

  • Infrastructure-as-Code generation
  • AWS architecture diagrams

Collaborative Development

  • Multi-user project sessions
  • AI-assisted design collaboration

New AI Agents

  • Security architecture agent
  • DevOps automation agent
  • Performance optimization agent

Voice-Driven Development

  • Integration with Amazon Nova Sonic
  • Conversational project creation workflow

Our long-term vision is an AI engineering partner capable of designing, building, testing, and deploying complete systems autonomously.

Built With

Share this project:

Updates