Author
Kim
Backend software engineer working with Java, Spring Boot, and Kafka. I write about engineering problems, performance work, and architecture decisions from development and operations.
-
Integrating Explainable AI Techniques into Production ML Workflows: A Comprehensive Guide
This guide helps ML engineers integrate SHAP explainability into production workflows, covering setup, verification, failure modes, and best practices for trust and compliance.
-
Implementing Effective Chaos Engineering Practices in Production Systems
Explore practical strategies and tools to implement effective chaos engineering practices that strengthen production system resiliency.
-
Implementing Kafka Tiered Storage for Cost-Effective Long-Term Data Retention
Discover a practical guide to Kafka tiered storage for scalable, affordable long-term data retention leveraging cloud object storage.
-
Using Spring Boot Batch for Large-Scale Data Processing with Step-by-Step Operational Guide
A step-by-step guide to building and managing Spring Boot Batch jobs for reliable, efficient large-scale data processing in enterprise environments.
-
Implementing Java Flight Recorder (JFR) for Real-Time Application Performance Monitoring
Implement Java Flight Recorder for real-time application monitoring with minimal overhead and deep JVM integration.
-
Optimizing Inference Latency for Deep Learning Models in Cloud Environments
A detailed guide for AI engineers on optimizing deep learning inference latency in cloud environments, covering model quantization, asynchronous inference, hardware choices, and production safeguards.
-
Implementing Robust Log Aggregation and Analysis for Scalable Systems
Learn to build and operate a fault-tolerant Fluentd and Elasticsearch log aggregation pipeline for scalable distributed systems with structured JSON logs and practical examples.
-
Kafka Consumer Groups Under Failure: What a Killed Consumer Costs with Range, Sticky, Cooperative and KIP-848 Assignment, Measured on Kafka 4.3.1
Three consumers, 48,000 records, one SIGKILL: 44 duplicates with RangeAssignor, 46 with StickyAssignor, 19 with CooperativeStickyAssignor, 15 with the KIP-848 protocol, and 0 with a graceful close. Why committing in onPartitionsRevoked never runs for a crash, and where the assignor lives now.
