Skip to content
AwakastLet's talk

ValidLabels, our own SaaS, built with AFL Laboratories

ValidLabels manages CLP labels for 1,200+ SKUs and prints them in under 7 seconds

Chemical safety labels have to be exact in the language of every market. Our own SaaS, built with the regulatory specialists at AFL Laboratories, translates them into 24+ EU languages, checks each phrase and sends the label straight to the packing-line printer.

From an idea

SKUs managed in the first production pilot
1,200+
from scan to printed label on the packing line
<7s
EU languages translated automatically
24+
lower cost per label than regulatory consultants
90%

Where it started

Under the EU's CLP Regulation, every hazard statement, precautionary phrase and pictogram on a chemical label has to be right in the language of each market where the product is sold. The required phrases also change with regulatory updates. Companies with portfolios of 1,200+ SKUs across 10+ EU markets relied on regulatory consultants, which was expensive and took days for every update. Doing it by hand meant high cost, high risk, or both. AFL Laboratories brought deep knowledge of CLP rules and chemical safety work. Together we set out to turn that knowledge into software.

What we built

We designed and shipped ValidLabels end to end as a multi-tenant SaaS. AI translation modules trained for CLP map each sentence to the official phrase set. A language model handles free-text safety descriptions within strict limits, and the result is checked before anything is saved. Every label has a full version history, and teams work with editor, validator and admin roles. Labels print directly on Zebra printers on the packing line. Everything runs on a self-hosted Kubernetes cluster in the EU.

Deliverables

  • Multi-tenant SaaS with team roles and single sign-on (OAuth 2.0)

  • AI modules that map each sentence to the official CLP phrase set

  • Language-model translation of free-text descriptions, checked before saving

  • Zebra printing (ZPL), under 7 seconds from scan to label

  • One repository for labels and safety data sheets, with full version history

  • Alerts on saved labels when a CLP update affects them

How it works

  1. Each sentence is matched, not just translated.

    AI modules map every hazard and precautionary sentence to the official CLP phrase for that language. A sentence that doesn't match is stopped at the validation step.

  2. Free text gets guardrails.

    A language model translates product descriptions with prompts limited to the CLP domain, and the result is checked before the label is saved.

  3. Every change has a name and a date.

    Each edit and approval is stored with a timestamp and the person who made it, so the team can show the full history of any label.

  4. Scan, print, pack.

    On the packing line, a scan sends the right label to a Zebra printer in under 7 seconds. Label data stays on EU-hosted infrastructure.

Results

SKUs managed in the ELiX pilot
1,200+
scan to print on the packing line
<7s
EU languages translated automatically
24+
lower cost than regulatory consultants
90%
  • CLP labels ready in minutes instead of days, without waiting for consultants

  • Printing straight from the packing line replaced manual label handling on the production floor

  • A full history of every label, for inspections and traceability questions

How we built it

  1. Phase 1

    Regulatory research

    Mapped CLP Regulation (EC) No 1272/2008, the hazard phrase taxonomy, the multilingual phrase annex and market-specific edge cases.

  2. Phase 2

    Platform and data model

    Built the multi-tenant architecture, the label repository, the versioned PostgreSQL schema, OAuth 2.0 and team permissions.

  3. Phase 3

    AI translation

    Built the CLP phrase-matching modules, the language model for free text with domain limits, and the full validation pipeline.

  4. Phase 4

    Printing and production pilot

    Built the Zebra ZPL output, tested timing on the packing line, ran the ELiX pilot with a 1,200+ SKU portfolio and hardened the platform for production.

Live SaaS product

Technology

  • Python
  • React
  • PostgreSQL
  • OAuth 2.0
  • Kubernetes
  • Custom AI modules
  • LLM
  • Zebra ZPL

Show us the expert knowledge you'd like to turn into a product.