Startup Application Review Agent
An AI-assisted application review system built on Azure AI Foundry, Fabric, Dataverse, PowerApps, and Azure AI Search for MFS and AFS application checks, duplicate detection, and approve / reject recommendations.
What was broken
Program and support teams reviewed startup applications across fragmented systems — investor codes, OneVet fields, duplicate applications, LinkedIn profiles, startup countries, and ticket resolution history. The process depended on manual checks across multiple sources, making the review slow, inconsistent, and hard to audit.
How I built it
I built an agentic review system with Fabric as the governed data layer and Azure AI Foundry as the agent layer. PowerApps gives reviewers an application status, ticket recommendation, and ticket details surface. Dataverse stores transactional data for Ticket Resolution and Application Review. External sources (DFM, FH CRM, SharePoint / Loop) land in a Fabric Lakehouse via Shortcut, Dataflow, and Pipeline on a Medallion architecture with backup and domain knowledge infusion. Two specialised Foundry agents — MFS Application Agent and AFS Application Agent — apply their own checks (investor code, OneVet, duplicates, LinkedIn, startup country, tech-based classification) and return an approve / reject recommendation. Vectorised ticket and domain knowledge in Azure AI Search grounds every agent call, refreshed on a daily cadence.
The reasoning
Splitting MFS and AFS into dedicated agents keeps each prompt focused on the rules that actually apply, with its own evals and retry boundary. Fabric + Medallion architecture gives a single governed data layer; Dataverse keeps the review state transactional; AI Search vector indexes give the agents grounded retrieval over tickets and domain knowledge instead of hallucinated rules.
How the pieces fit
Key components
- 01PowerApps reviewer UI for application status and ticket recommendations
- 02Dataverse transactional storage for tickets and review state
- 03Fabric Lakehouse with Medallion architecture, Dataflows, Pipelines, and backups
- 04MFS agent: investor code, OneVet fields, duplicate checks, approve / reject
- 05AFS agent: LinkedIn profile, startup country, tech-based classification, duplicates, approve / reject
- 06Vectorised domain knowledge + structured ticket data indexed in Azure AI Search
- 07Daily ingestion / refresh cadence from source systems into Fabric and vector knowledge
- 08Power Automate flows wiring PowerApps actions to Dataverse and the agent layer
Tech stack
In-browser demo
A scripted, in-browser walkthrough of a real run, traced step by step. Press play to watch the agents fire.