Industry Served: EdTech
Druid Learning, Turning Decades of Content Archives Into AI-Ready Data
An End-to-End Content Intelligence Platform for Publishers, Broadcasters, and Research Institutions
From Content Mess to Usable AI Data
Discuss your projectDruid Learning is a Dublin-founded technology company that transforms decades of archives, content libraries, and information repositories into AI-ready data. Publishers, broadcasters, and research institutions sit on enormous proprietary content estates: backlist titles, scanned books, audio-visual files, legacy flash material, PDFs, and metadata scattered across a dozen systems. Very few have the infrastructure to safely optimise and capitalise on it.
We developed the entire platform: a data intelligence layer that ingests content from any source, uses machine learning and natural language processing to enrich and standardise its metadata, and activates the resulting proprietary dataset inside the customer's existing analytics, editorial, and AI applications. Crucially, the source content is never altered and data ownership never leaves the client.
The platform now serves organisations including NPR, CJ Fallon, the Dutch Cancer Institute, and the University of Bristol, and is backed by Techstars, Enterprise Ireland, the European Union's Horizon programme, and LvlUp Ventures. Druid Learning was named winner of Frankfurter Buchmesse's 2026 Wildcard.
Project Challenges & Needs
Fragmented, Multi Format Archives
Content sat on FTP servers, in legacy databases, and behind APIs, as OCR, XML, INDD, MP4, PDF, EPUB, and legacy flash files.
Manual Micro-Categorisation
Itemising blocks of digital content and tagging component assets was manual, slow, and expensive to manage at scale.
Inherited, Inconsistent Metadata
Libraries acquired through acquisition or built on author-submitted records had no consistent, usable schema.
No Way to Break Files Apart
A single PDF could hold images, text, tables, and links with no means of separating them into individually addressable, reusable assets.
Our Solution
We built a scalable, modular Content Intelligence Platform enabling:
The platform is delivered as connected products sitting on one pipeline, so clients can start by indexing and categorising an archive and expand into analytics, editorial suggestions, and metadata standardisation without changing their IT infrastructure or existing content.
Interfaces We Developed
Druid Learning Administrator Interface
The highest level of administration, allowing the Druid Learning team to:
- Add, edit, delete, activate, and deactivate every client organisation on the platform
- Monitor indexing and categorisation volume across all client accounts
- Manage taxonomies, parameter layers, and platform-wide configuration
Client Administration Interface
A workspace for each publishing house, university, or institution, enabling client authorities to:
- Add, edit, delete, activate, & deactivate their users
- Define proprietary categories, taxonomies, and labels for their content estate
- Review AI-generated tags and validate sample datasets before scaling up
- Track saved projects with created date, indexed files, and total assets
Content Creator Interface (Indexing & Categorisation)
The core working environment, built around an Index menu covering Add File, Indexing History, Saved Projects, and Create New Project. Content creators can:
- Upload files singly or in bulk, with type, page count, user, version, and date captured on ingest
- Index a whole file, every uploaded file at once, or a chosen page range
- Review output showing total assets and a breakdown by image, text, table, and link
- Browse assets by type or view every asset from a file on one page
- Send any asset through the AI categorisation flow
- File assets into folders defined by topic area, level or grade, and subject
- Build projects from indexed files, assets, and whole folders
- Export any asset, folder, or project to local storage
Key Features & Capabilities
Centre & Course Setup
Admin defines courses, levels, sessions, and onboards each learning centre and DOS.
AI-Powered Micro-Categorisation
Machine learning labels every file by content topic, cross-labels for relevancy, and semantically links related content across the archive.
Custom Page-Range Indexing
Creators index a chosen range of pages rather than reprocessing entire documents, with full indexing history retained.
Projects, Folders, and Export
Assets organised into projects and folders by topic area, level or grade, and subject, then exported at asset, folder, or project level.
Human-in-the-Loop Validation
Sample datasets validated with subject-matter experts before scaling, built into the process rather than bolted on.
Client-Owned Data
Source content is never altered and clients retain full ownership and control of their dataset.
Outcomes and Impact
500k+
Assets Processed
100k+
Data Points Generated
140+
Languages Supported
100%
Of Your Data, Your Contro
NPR
Public Broadcasting
Audio, video, and editorial archives consolidated and transformed into a labelled, AI ready dataset. 500k+ assets processed.
Dutch Cancer Institute
Medical Research
3,000 research papers enriched, yielding hundreds of thousands of structured data points.
CJ Fallon
Education Publishing
Unified metadata and curriculum-mapped archive across the full primary-school content library. 100% curriculum coverage.
Technologies Used to Develop the Solution
Web Frontend
Next.js
Admin, client, and content creator interfaces for indexing, projects, and asset management.
Backend
Python (FastAPI) & Node.js
Ingestion, indexing, categorisation, and delivery APIs across the pipeline.
AI, Machine Learning
Transformer Models & Classifiers
Automated topic labelling, relevancy cross-labelling, and content classification at scale.
Natural Language Processing
spaCy & Named Entity Recognition
Extracts people, dates, and topics, and supports processing across 140+ languages.
Content Extraction
OCR & PDF Parsing
Separates text, images, tables, and links out of PDFs, scans, and legacy formats.
Data Pipeline
Apache Airflow & Kafka
Orchestrated, resumable processing of very large archive batches.
Database & Search
PostgreSQL & Elastic search
Structured metadata storage with fast search across millions of assets.
Integration
GA4, Chart beat, Tableau, WordPress, Salesforce, Airtable
Enriched data delivered into editorial, archival, and analytics tools already in use.
Cloud Infrastructure
AWS & Google Cloud
Elastic compute for large batch processing, with client-controlled data residency.
Security
Role-Based Access & Encryption
Separates Druid Learning, client, and user permissions and protects proprietary content end to end.
Here's What Our Client Has to Say
The team took the time to understand a genuinely complex problem before writing a line of code, and that showed in what they delivered. They built the indexing and categorisation engine our product depends on, stayed with us through several phases of development, and handed over clean documentation and full ownership at every stage. They have been a dependable technical partner as our business has grown.
Get in touch
Share a few details about what you're building, and we'll take it from there.
