https://kozlov.ski/This is my cool siteHugo -- gohugo.ioennkozlowski@pm.me (Norbert Kozlowski)nkozlowski@pm.me (Norbert Kozlowski)Tue, 03 Mar 2026 08:00:00 +0100https://kozlov.ski/demand-forecast-eda/Wed, 25 Feb 2026 08:00:00 +0100Norberthttps://kozlov.ski/demand-forecast-eda/A client handed me 1.5M rows of POS data and asked for daily demand forecasts. EDA revealed 66% of series were lumpy and unforecastable — before we trained a single model. Here’s the complete analysis from a data engineer with 15 years of experience and a Ph.D. in AI.https://kozlov.ski/activity-schema/Wed, 04 Jun 2025 08:00:00 +0100Norberthttps://kozlov.ski/activity-schema/One table. Temporal joins. No foreign keys. Activity Schema is a radical simplification for event analytics - here’s when it makes sense and what tradeoffs to expect.https://kozlov.ski/decision-making-under-uncertainty/Mon, 30 Sep 2024 08:00:00 +0100Norberthttps://kozlov.ski/decision-making-under-uncertainty/When should you explore vs. exploit? Thompson Sampling offers an elegant Bayesian answer to this fundamental trade-off. A practical framework for A/B testing, dynamic pricing, and any decision where you’re balancing learning against optimization.https://kozlov.ski/mautic-5x-kubernetes-setup/Wed, 04 Sep 2024 13:00:00 +0100Norberthttps://kozlov.ski/mautic-5x-kubernetes-setup/A step-by-step guide to deploying Mautic 5.x on Kubernetes with Helm, including the background job architecture data engineers need to understand for reliable integrations.https://kozlov.ski/developers-guide-to-data-lakehouse-with-apache-iceberg/Fri, 22 Mar 2024 13:00:00 +0100Norberthttps://kozlov.ski/developers-guide-to-data-lakehouse-with-apache-iceberg/A hands-on guide to building a Data Lakehouse with Apache Iceberg. Covers table specifications, atomic transactions, schema evolution, hidden partitioning, time travel queries, and maintenance procedures using Spark and Trino with a local MinIO setup.https://kozlov.ski/budget-data-lake-with-aws/Fri, 02 Feb 2024 12:05:02 +0200Norberthttps://kozlov.ski/budget-data-lake-with-aws/Build a serverless, cost-efficient data lake on AWS using Kinesis Firehose, S3, Glue, and Athena. Learn dynamic partitioning, JSON vs Parquet trade-offs, and query optimization techniques that keep your AWS bill under a dollar.