ASA ( Autonomous Solutions Architect )

Discover. Design. Deliver.

Comment

ASA (Autonomous Solutions Architect)

Inspiration

Cloud discovery workshops are one of the most critical phases of any cloud adoption journey, yet they remain largely manual, consultant-driven, and time-intensive. Organizations often spend days or weeks gathering requirements, validating assumptions, and producing architecture recommendations before implementation can even begin.

I wanted to explore whether an autonomous AI could take on the role of a Principal Solutions Architect—leading the discovery workshop, asking intelligent follow-up questions, reasoning through business and technical requirements, validating its own recommendations, and producing implementation-ready deliverables. Rather than building another AI chatbot, my goal was to create an AI consultant capable of owning an entire cloud discovery engagement.

What it does

ASA (Autonomous Solutions Architect) is an AWS-native, multi-agent AI platform that conducts cloud architecture discovery workshops through natural voice and chat conversations.

During a workshop, ASA continuously reasons about the customer's objectives, identifies missing information, plans the next steps, and coordinates specialist AI agents to complete the engagement. Instead of following a fixed questionnaire, it dynamically adapts the conversation based on the customer's responses.

The platform collaborates across multiple specialist agents responsible for discovery, architecture design, planning, and independent review before generating implementation-ready consulting deliverables such as:

  • Executive Summary
  • Business & Technical Requirements
  • AWS Architecture Recommendations
  • Architecture Decision Records (ADR)
  • Mermaid Architecture Diagrams
  • Risk Register
  • Migration Roadmap
  • Markdown and PDF Reports

How we built it

ASA was designed as a modular, AWS-native, multi-agent system following an autonomous reasoning loop:

Observe → Reason → Plan → Select Agent → Invoke Tools → Evaluate → Reflect → Repeat

The solution leverages AWS-native technologies including:

  • Amazon Bedrock for generative AI capabilities
  • Amazon Nova Sonic for natural voice conversations
  • Amazon Bedrock AgentCore for agent execution
  • FastAPI for backend APIs
  • React + TypeScript + Tailwind CSS for the user interface
  • Amazon API Gateway for API communication
  • Amazon DynamoDB for persistent session memory
  • Amazon S3 for report and artifact storage
  • AWS Lambda for modular agent execution
  • AWS Step Functions for orchestration workflows
  • Terraform for Infrastructure as Code
  • GitHub Actions for CI/CD

The platform separates planning from execution using a Planner and Orchestrator architecture, allowing specialist agents to collaborate independently while maintaining structured memory and reflection before producing final recommendations.

Challenges we ran into

Designing an application that genuinely demonstrates agentic behavior proved to be the biggest challenge. It was tempting to implement a traditional chatbot with a series of prompts, but that would not satisfy the project's vision.

Another challenge was designing clear boundaries between reasoning and execution. Separating the Planner from the Orchestrator required careful architectural decisions to avoid creating a monolithic system where one component handled every responsibility.

Balancing hackathon constraints with production-ready design was also challenging. I had to prioritize the capabilities that best demonstrated autonomous reasoning while keeping the implementation achievable within a limited timeframe.

Accomplishments that we're proud of

The accomplishment I'm most proud of is successfully designing ASA as an autonomous AI consultant rather than a conversational assistant.

Key achievements include:

  • Designing a true multi-agent architecture with clearly defined responsibilities.
  • Implementing a planning model that dynamically determines the next step instead of relying on scripted workflows.
  • Introducing an independent Review Agent that validates architecture recommendations before they are presented.
  • Building a structured memory model that preserves requirements, assumptions, risks, and architectural decisions throughout the engagement.
  • Creating a scalable architecture that can easily accommodate additional specialist agents and enterprise integrations in the future.

What we learned

This project reinforced that building agentic AI requires much more than connecting an LLM to a user interface.

True autonomy comes from combining reasoning, planning, memory, tool usage, reflection, and collaboration between specialized agents.

I also gained a deeper appreciation for designing systems around business capabilities rather than individual cloud services. By focusing on agent responsibilities and clean interfaces first, the underlying AWS implementation became significantly more modular and extensible.

Finally, I learned that voice-first interactions create a much more natural discovery experience when paired with structured planning and contextual memory.

What's next for ASA (Autonomous Solutions Architect)

This project represents the foundation of a much broader vision for AI-powered cloud consulting.

Future enhancements include:

  • Support for Azure, Google Cloud, and hybrid cloud architectures.
  • Live AWS pricing analysis and cost optimization recommendations.
  • Automated Terraform and AWS CloudFormation generation.
  • Direct integration with AWS Well-Architected Tool and AWS Architecture Center.
  • Retrieval-Augmented Generation (RAG) using AWS documentation, reference architectures, and best practices.
  • Multi-session organizational memory for returning customers.
  • Collaborative workshops involving multiple stakeholders simultaneously.
  • Integration with Jira, GitHub, Confluence, Slack, and Microsoft Teams.
  • Automated implementation roadmaps and migration planning.
  • Compliance-focused specialist agents for industries such as healthcare, finance, and government.

Ultimately, the vision for ASA is to become an autonomous Autonomous Solutions Architect capable of partnering with organizations throughout their cloud adoption journey - from the first discovery conversation to implementation planning and continuous architectural improvement.

Built With

Share this project:

Updates