Snowflake User Group Ireland Meetup

Curious about what Irish SMEs are actually doing with AI? We’re getting together with the folks from The Irish Times, Buymedia, and Coolplanet to look under the hood of their data stacks. They'll be sharing exactly how they use Snowflake AI to clean up messy documents, speed up data engineering, and turn weeks of work into hours. Drop by for some honest talk about what works in the real world, plus food, drinks, and a chance to catch up with the local Snowflake community.

Right Story, Right Reader: Building a Hybrid News Recommendation Engine in Snowflake, Juliana Murphy, The Irish Times

At The Irish Times Group — home to The Irish Times, Irish Examiner, Breaking News, The Echo, and more — we publish hundreds of articles every week. The challenge isn’t the quality of the journalism, but helping readers find the stories most relevant to them, as discovery becomes increasingly fragmented across search, social, and other channels.
In this session, Juliana Murphy walks through how The Irish Times built For Me — a pilot personalised content feed delivered three times daily to subscribers — using a fully native Snowflake architecture, with no external ML infrastructure or data movement outside the warehouse.
The session covers the full pipeline: converting article text into 1,024-dimensional vector embeddings with Snowflake Cortex, building subscriber preference profiles from behavioural data, and using native vector similarity scoring and task orchestration to generate recommendations end-to-end within the warehouse, with output delivered to subscribers via an S3 stage.
Beyond the technical build, we’ll explore the constraints unique to news. Unlike entertainment platforms, journalism must balance personal relevance with information diversity. We’ll show how editorial curation, controlled ranking logic, and user choice are embedded directly into the system design — and how those decisions shape both architecture and outcomes.

Key takeaways:

• Why editorial and ethical constraints belong in your data architecture, not just your product spec
• Building a production-grade recommendation pipeline natively in Snowflake (Tasks, Cortex embeddings, cosine similarity scoring, Python UDFs)
• What “hybrid” means in practice: combining ML signals with editorial control
• Why domain knowledge matters as much as model performance

Accelerating Data Engineering with Snowflake Cortex Code, Jenny Zhou, Snowflake

This demo showcases Cortex Code, Snowflake’s AI-powered coding agent designed to automate and optimize the end-to-end data development lifecycle. We will demonstrate how Cortex Code moves beyond simple code completion by leveraging platform awareness—integrating your database schema, security context, and execution logs directly into the development workflow.

Extracting Structure from Chaos: Parsing Documents from 400+ Media Sources with Cortex AI at Buymedia, Estella Roberts, Buymedia

At Buymedia, ingesting data from over 400 unique media sources is anything but clean. Every source brings its own document formats, layout quirks, and edge cases that completely break traditional extraction tools.

In this session, Estella Roberts shares how Buymedia transformed a manual, error-prone bottleneck into an automated, native pipeline using Snowflake Cortex AI and CORTEX.COMPLETE.
We'll dive deep into the architecture, explore the messiest edge cases, and share practical solutions for when unstructured documents refuse to cooperate.

Expect less theory, more real-world trade-offs, and actionable lessons on building resilient, modern AI pipelines.

From Sprints to Hours: Building an AI-Powered Data Stack at Coolplanet, Niall Magee, Coolplanet

At Coolplanet, we handle massive volumes of customer data effortlessly. But our internal business data? That started with Google Sheets, a basic BI tool, and a quiet confidence that it would scale. It didn't.
In this session, we’ll share our honest journey from a fragile spreadsheet setup to a production-grade modern data stack using Fivetran, dbt, and Snowflake. More importantly, we'll reveal how AI became our ultimate force multiplier, turning sprint-long tasks into hours-long workflows.
The highlight of the session is a deep dive into our AI-driven build process. We’ll show you how to take a plain-language brief and immediately generate dbt models, semantic layers, and functional Streamlit dashboards.

Key Takeaways:

• The AI Workflow: Practical steps to accelerate dbt, YAML, and SQL generation.
• The Stack Evolution: Honest lessons from migrating to Fivetran, dbt, and Snowflake.
• Beyond Traditional BI: Why we are shifting toward AI-scaffolded Streamlit apps.
• Snowflake Cortex: How to ground natural language queries in a governed semantic model.

Location
Clayton Hotel Charlemont

Agenda

6:00 pm – Food & Networking
6:30 pm – Welcome
6:35 pm – Right Story, Right Reader: Building a Hybrid News Recommendation Engine in Snowflake