AI Integration Lab

An interview-ready portfolio project for an AI Integration Specialist role in a multi-country enterprise.

The demo models the work described in the role: discover a department process, identify an AI opportunity, validate economics, apply data-safety gates, plan UAT/adoption and publish a Notion-ready decision record.

What it demonstrates

• Structured process intake with typed validation.
• Explainable matching of business needs to an internal AI solution catalog.
• ROI, monthly net benefit and payback calculations.
• Data-classification gates for internal, confidential, personal and health data.
• Human-in-the-loop controls, RBAC, auditability and rollout criteria.
• UAT scenarios, adoption targets and success metrics.
• A small FastAPI API plus a no-build browser UI.

Run locally

cd ai-integration-lab
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -e ".[dev]"
uvicorn app.main:app --reload

Open http://127.0.0.1:8000.

Run tests:

pytest

API

• GET /api/health
• GET /api/catalog
• GET /api/demo-case
• POST /api/audit

The app is deterministic by design: it runs without API keys and makes the decision logic inspectable during an interview. An LLM provider can later be added behind the same structured input/output boundary for summarisation or process discovery, while ROI and security gates remain deterministic.

Demo story

Use the preloaded Sales Operations case. The system identifies AI Document Intake & Validation as the best pilot, estimates economics, lists controls and produces a Notion-ready brief. Then change data classification to health_data or choose a generic public chatbot scenario to show that the safety gate can block an attractive but unsafe option.