AI-103 · In Progress

Intelligent Insights Platform

A Document Intelligence System built on Azure AI Foundry — OCR and structured extraction via Content Understanding, RAG-grounded summarization and Q&A over a real vector index, responsible-AI evaluation, and a multi-agent orchestration pattern. Built while studying for AI-103 (Developing AI Apps and Agents on Azure), the associate-level follow-on to AZ-900 and AZ-104, both already earned.

6 / 7Milestones complete

M7Currently in progress

2Certs already earned

What it does

Upload a technical document — a clean PDF export, a flatbed scan, or an angled phone photo — and the pipeline extracts text and structured fields, answers natural-language questions grounded in the document's actual content (not the model's memory), and checks its own output for fabrication before returning it.

Extraction Azure Content Understanding — structured JSON with confidence scores and source grounding, validated against clean PDF, flatbed scan, and angled-photo conditions of the same document.
RAG / Q&A Chunked, embedded, and indexed into Azure AI Search; queries retrieve and join multiple ranked chunks (not just the top hit) before grounding a chat completion — a design choice a live test proved actually matters.
Evaluation A custom evaluator harness (Groundedness / Relevance / F1) used to pick the production model on evidence, not assumption — confirmed a real ~3x cost advantage at equal quality.
Agents (M7, not yet built) A single orchestrator agent drafting content against a template, reusing the M6 evaluator pattern for QA, plus a computer-vision audit module.

Deliberately out of scope: image/video generation and speech modalities — poor fit for a document-intelligence use case. Covered through study material for the exam, not the build. A documented choice, not a gap.

Milestones

Complete

Tore down the original West US 3 resource group and all resources — zero carryover into the rebuild.

Complete

Redeployed Azure AI Foundry fresh in West US, with quota re-verified against the new region before building on it.

Complete

Custom document analyzer built and validated against the clean-PDF loan agreement sample document.

Complete

Submit-and-poll extraction pipeline written, debugged, and validated end-to-end against the analyzer.

Complete

Extraction validated across all three image-quality conditions — clean PDF export, flatbed scan, and an angled phone photo of the same document.

Complete

RAG-grounded Q&A working end-to-end against a real Azure AI Search vector index — live-verified with a retrieval-ranking finding: the top-ranked chunk isn't always the correct one, so the pipeline joins every retrieved chunk into context rather than trusting rank alone.

Complete

Evaluator harness built and run twice against independent question sets; selected the production model on confirmed quality parity plus a real ~3x per-token cost advantage, not on assumption.

Not started

Single orchestrator agent over a synthetic content-review scenario — drafts against a template, reuses the M6 evaluator pattern for QA, and adds a computer-vision audit module covering the one AI-103 exam domain not otherwise touched. Closes out Phase 1 of the IIP build.