logologo
AboutResumeProjectsField Notes

© 2026 Beretests. All rights reserved.

← Back to projects
Power PlatformCompleted

Public project

Power Platform Architect Agent

Solution architect and full-stack engineer

Designed and built an AI-assisted architecture workspace that converts Power Platform requirements and existing designs into structured solution blueprints, review findings, readiness scores and implementation guidance.

Power PlatformNext.jsTypeScriptAzure OpenAIFoundry IQAzure AI SearchZodMermaid
Watch demoView GitHub repository

What the work delivered

01

Produces consistent architecture outputs covering Dataverse, Power Automate, security, ALM and risk

02

Makes grounding sources and AI limitations visible to reviewers

03

Exports implementation-ready Markdown, JSON and Mermaid artifacts

Goals

  • Accelerate early Power Platform architecture and design reviews
  • Surface governance, security and production-readiness gaps before delivery
  • Provide predictable, reviewable outputs instead of unstructured AI responses

Key design decisions

  • Used Azure OpenAI structured output with Zod validation for predictable rendering
  • Grounded responses with Foundry IQ and Azure AI Search, with curated local guidance as a fallback
  • Separated architecture generation and Solution Review Board modes
  • Included source transparency, responsible-AI notices and a clearly labelled demo mode

Challenges and solutions

Reliable AI output

Defined strict architecture and review schemas, requested JSON-schema output and validated every response before rendering it.

Trustworthy grounding

Added Foundry IQ retrieval, local knowledge fallback and visible grounding metadata so users can distinguish sourced guidance from demo data.

Key learnings

  • Designing AI features around human review rather than automatic approval
  • Combining retrieval grounding with structured outputs for enterprise use cases
  • Translating Power Platform governance standards into reusable architecture checks