Databricks NL Community Night – MLOps Edition
Description
Running ML and GenAI systems in production is hard - especially once experiments turn into real, business-critical systems.
This Databricks Community NL Night brings together engineers and practitioners who are building, operating and scaling production ML and AI systems on or around Databricks.
The evening will focus on:
ML Ops, GenAI and LLM Ops in practice
Architectural choices, trade-offs and lessons learned
What actually works (and what doesn’t) once models go live
Moving from experimentation to reliable production systems
The format is community-first and technical:
1–2 in-depth talks, followed by drinks, pizza and plenty of time to chat and connect.
Confirmed speakers
Oleksandra Bovkun
Senior Developer Advocate @ Databricks
Talk: Building High-Quality AI Agents: from Experimentation to Production with MLflow
Abstract:
While prototyping a Generative AI agent is straightforward, the true hurdle lies in ensuring consistent, high-quality behavior at scale. This session dives into the agent-driven development cycle, focusing on how to transition from experiment to production using MLflow. We will demonstrate how to leverage MLflow GenAI components to operationalize workflows and use MLflow scorers to rigorously evaluate agent performance, ensuring your deployment is both reliable and production-ready.
Eneco
Talk: ML Systems for Forecasting Long-Term Energy Consumption
Abstract:
Accurately forecasting energy demand over long horizons isn’t just a modeling challenge – it’s a full-stack engineering feat. This talk explores how we built a production-scale machine learning system for long-term energy consumption forecasting, addressing real-world challenges from data integration to model deployment. We discuss data interoperability across multiple platforms (Snowflake, Azure, Databricks), showing how we unified siloed data warehouses into a seamless pipeline that feeds our forecasts. Next, we outline the ML infrastructure and MLOps tools that power our solution – including the cloud architecture, workflow orchestration, and experiment tracking (using MLflow) – and explain how platform constraints shaped our design. The session will delve into modeling choices and trade-offs we made, such as why we adopted an ensemble of sub-models and how we navigated limitations of MLflow’s model management. Finally, we highlight our model monitoring, experimentation, and backtesting framework that ensures the deployed forecasts remain accurate and reliable over time. Attendees will learn practical strategies for deploying ML forecasting models in production, from handling messy data and tooling limits to maintaining performance through continuous monitoring and validation.
Who should attend
ML engineers and data engineers working with Databricks in production
Engineers building or operating ML, GenAI or LLM-powered systems
Platform or infrastructure engineers supporting ML and AI workloads
Technical practitioners interested in ML Ops, GenAI or LLM Ops in practice
Practical info
📍 Location: Kruispunt Event Space. TT Vasumweg 44, 1033 SC, Amsterdam
📅 Date: Thursday, April 2nd
🕦 Time: 6-9pm
🎟 Tickets: Free (limited capacity, waitlist enabled)
🍕 Food & drinks: Pizza, beers and non-alcoholic options provided
This is a community-led event hosted by RevoData and is not an official Databricks event.