ALL SPEAKERS
SESSIONS
SCHEDULE TBD
KEYNOTE
We built it wrong (on purpose): The Evolution of adjoe’s Feature Store
AT A GLANCE
Role
Senior Data Engineer
Organisation
adjoe
Category
Community
ABOUT THE SPEAKER
Luan is a Senior Data Engineer at adjoe, building the data infrastructure behind its ML platform. Bridging data science and backend engineering, he turns raw event streams into 600+ features driving ad relevance and recommendations. His work spans batch and real-time pipelines, feature store architecture, and data quality at scale. His stack includes Apache Kafka, Flink, dbt, Airflow, Druid, ClickHouse, and Kubernetes on AWS.
TALK
We built it wrong (on purpose): The Evolution of adjoe’s Feature Store
Feature stores are rarely built right the first time. They're grown, patched together from existing infrastructure, extended as model complexity increases, and eventually rebuilt when the original design hits its limits. This talk traces the full evolution of adjoe's internal feature store: from repurposed infrastructure and a handful of model features, to 600+ features powered by custom Go code with DynamoDB and dbt on Trino and Athena, to a near real-time processing system built on Apache Flink handling terabytes of data daily. The session covers the practical decisions made at each stage: what worked, what broke, and why the jump to Flink was ultimately necessary, offering a realistic blueprint for teams navigating the same trajectory.
MORE VOICES
Other speakers at code.talks
See them live — and 100+ more.
One ticket, every track. Nov 4–5, 2026 at Kinopolis Hamburg.
Get your ticket





