DataTribe Meetup #8 - Data Engineering and AI

# DataTribe Meetup #8 - Data engineering and AI @ Brightly

📅 Date: Thursday, August 27th, 2026
📍 Location: Brightly Office, Pohjoisesplanadi 37 A, 00100 Helsinki
⏰ Time: 18:00 – 20:30
🎯 Theme: Workflow improvements using AI, Offline models

🛠️ Agenda

• 18:00 – 👋 Welcome
• 18:20 - 🎤 DataTribe and Brightly introduction
• 18:30 – 🎤 Sanket Joshi, From Prototype to Production - Building Production-Ready AI Applications
• 19:15 – 🎤 Dan Suman, Engineering Data Agent Loops with offline LLMs
• 20:00 – 🍕🍻 Networking, Snacks, Drinks & Open Networking
• 20.30 - Doors close

From Prototype to Production - Building Production-Ready AI Applications by Sanket Joshi

• Sanket is also a Microsoft Certified Trainer and active tech community speaker. His focus is on the operational side of GenAI. The idea behind his talk is to show how teams can take an AI application from a working prototype to a production-ready system.
• In his talk, "Building Production-Ready AI Applications," Sanket will share:

🔹 Prompt management, evaluation, and guardrails for LLM outputs
🔹 Observability and tracing with tools like Langfuse
🔹 Monitoring, cost optimization, and a cloud-agnostic reference architecture for production deployment

Drawing on hands-on experience deploying AI systems across cloud and on-prem environments, Sanket will offer a practical, end-to-end architecture for anyone trying to move an AI proof-of-concept into something reliable and scalable.

Engineering Data Agent Loops with offline LLMs by Dan Suman, Lead Data Engineer at Rovio Entertainment.

• With ten years of experience building scalable distributed systems and real-time data pipelines, Dan has spent the past year deep in agentic workflows - building his own local alternative to tools like n8n, powered by self-healing prompt optimization with DSPy and open models like Qwen.
• In his talk, "Engineering Data Agent Loops with Offline LLMs," Suman will show teams how to turn recurring data engineering work into local, evaluation-driven workflows. He'll cover:

🔹 How Codex works as an agentic harness for open models like Qwen 3.8-27B, with DSPy optimizing prompts from workflow feedback
🔹 Workflow design, using dbt-based ETL and ML feature engineering as examples
🔹 Extracting evaluation datasets from real runs, self-improvement loops, and optimizing the cost of repetitive tasks

📌 Event Info

• Venue: Brightly Helsinki Office
• Food & Drinks: Dinner and drinks provided
• Audience: Software, data and AI professionals
• Community Partners: Brightly
• Note: Photos may be taken for recap & marketing. Attendance implies consent, always used respectfully. 👉 Join the tribe: datatribecollective.com