🐨 Koala

Pre-check EU AI Act compliance, catalog AI systems, and get regulatory answers.

#ai#tool#eu#regulation#compliancy#ai-systems#local-llm#ai-act

View Demo View Code

Overview

Koala provides a pre-check for EU AI Act compliance, helps catalog AI systems, and answers regulatory questions.

Problem

The AI Act is complex and fast-moving, making early compliance checks difficult for product teams.

Target Users

Product managers and founders
Compliance and legal stakeholders
ML and engineering teams

Goals

Provide a quick compliance pre-check
Reduce uncertainty around regulatory requirements
Make policy questions easier to answer

Non-Goals

Automated policy drafting
Continuous monitoring of model changes
Organization-wide inventory admin

Solution

A guided assessment and Q&A flow tailored to AI Act requirements.

Koala overview dashboard

My Role

Personal case study. I translated the AI Act into a product flow by mapping obligations to a step-by-step intake, defined the assessment logic, and built the demo.

Key Decisions & Trade-offs

Led with a guided questionnaire to make inputs consistent; this meant giving up free-form document uploads and deeper nuance.
Prioritized actionable output over exhaustive legal detail.
Kept the flow lightweight so it can be adapted to different AI systems.

Key Features

Catalog intake (AI system entries)
Role-aware risk and obligation summaries
Regulatory Q&A with citations

AI Use

LLM used for regulatory Q&A and to interpret catalog entries.
Outputs are advisory and do not replace legal review.

Risks & Constraints

Legal interpretation still requires expert review
Regulatory updates may require prompt refreshes
Constraint: The AI Act is evolving, so I framed this as a pre-check rather than a compliance guarantee.

What I’d Do Next

If I restarted, I would design the AI system inventory schema first and build guidance around it.
Add an exportable compliance memo per system.
Provide tailored checklists by AI system type.
Maintain a regulation update log with change highlights.

Status / Deployment

Showcase demo. Demonstrates what is possible after self-deployment. Public demo is scripted; full ingestion and RAG run locally.