DueWeekly Tech: 16-06-2025 to 29-06-2025

AIOps / MLOps / LLMOps

1. Medium - Why Most AI Agents Fail in Production (And How to Build Ones That Don’t)

• At first glance, this is a particularly interesting article that comprehensively covers the information needed to build and succeed with AI Agents—one of the most significant and intriguing topics of recent months. You’ll find a checklist and a step-by-step guideline for approaching AI Agents and building them from scratch.
• Through each step, you’ll gain insights into necessary actions, recommended tools and resources, and how to ensure the success of your AI Agent by the end of its development journey.

Architecture

1. Datacamp - Database Sharding: Examples, Strategies, Tools, and More 🌟 (Recommended)

• This article defines Database Sharding, explaining when and why it should be implemented. It also helps distinguish between Sharding and Partitioning, highlighting their conceptual similarities but distinct meanings. Furthermore, the article elaborates on the differences between Sharding and Replication.
• It then explains how database sharding works, details various sharding strategies, and lists tools that support sharding for MySQL, MongoDB, and PostgreSQL.

2. Medium - ⚡️CacheBolt: Building a High-Performance Internal Cache Proxy That Actually Scales

• This article introduces CacheBolt (https://github.com/msalinas92/CacheBolt)—a new open-source CDN designed to provide internal caching. Its configuration can significantly enhance the latency of internal traffic. The author emphasizes that CacheBolt outperforms Redis in several preparation steps, requiring only a single config.yaml file.
• CacheBolt offers a few intriguing customization features:
  1. Latency: Define the maximum latency for each routing path.
  2. Memory Eviction: Set thresholds for when CacheBolt should evict data, preventing blind crashes.
  3. Multiple Cloud and Local Storage Support: It supports various storage options, including s3, blob, gcs, and local disk.
• Finally, the author provides a complete config.yaml file, allowing you to quickly review and easily configure what you need.

Career / Story

1. Medium - DevOps, MLOps, or Platform Engineering, In 2025, who will own the pipeline?

• Click the button to read about one of the biggest confusions in a DevOps career: how the position is often assigned multiple tasks across various fields, sometimes leading to unclear and dramatic situations. 😄
• In this article, the author discusses the distinctions between these roles and emphasizes that DevOps is the fundamental root from which other positions have evolved today. However, boundaries should always be set to ensure consistency and higher isolation within responsibilities.
• The article highlights that pipelines are no longer solely the domain of DevOps Engineers; they are a shared responsibility, with everyone in a product team contributing to different pipeline categories. As the author states: “Small changes can propel a business to immense success, so why not consider building huge success one small piece at a time?

Data Engineer

1. Medium - A Data Scientist’s Guide to Data Streaming 🌟 (Recommended)

• This article offers a unique perspective on streaming pipelines, specifically from a Data Scientist’s viewpoint. You’ll gain insight into their definition and the compelling reasons why streaming pipelines are so crucial today. The piece highlights why streaming often holds more significance than batch processing, underscoring that while both are valuable, streaming is sometimes the necessary approach.
• Tools are essential, and as expected, Flink and Kafka are prominently featured. Their combination forms the core of the streaming process discussed. The article connects these technologies to the architectures of major companies like Netflix and Pinterest, illustrating why these industry leaders adopt this particular pattern.
• Finally, the author presents real-world scenarios and code examples using PyFlink to solve specific problems. You’ll see how GPT-4 can handle complex NLP tasks, making this a fantastic topic for beginners and a great starting point for further exploration.

2. Medium - 8 Awesome Docker Images Every Data Engineer Needs on Their Radar in 2025

• If you’re a data engineer working with numerous tools, Docker is undeniably your friendly companion, no 🧢. It offers a wealth of features for portability, reproducibility, and isolation. Given the complexity, packaging individual tech stacks into Docker containers is a crucial practice for data engineers, allowing them to separate complex components into smaller, manageable pieces.
• This article provides a comprehensive collection of resources. You can refer to it anytime you need a solution for self-hosting a local environment for data engineers to begin experiments, covering tools like Airflow, Spark, Kafka, AirByte, and more.

Kubernetes

1. Medium - How We Load Test Argo CD at Scale: 1,000 vClusters with GitOps on Kubernetes 🌟 (Recommended)

• This article focuses on stress-testing Argo CD with large clusters, experimenting with Kubernetes multi-tenancy via vClusters. It provides benchmark results and explores how to efficiently scale Argo CD. The author also mentions Kubara, a templating binary that helps bootstrap a secure, GitOps-enabled Kubernetes control plane, as discussed in the LF Live Webinar: From CI to Kubernetes Catalog: A Practical Guide to vCluster and GitOps video by The Linux Foundation.
• The article walks through and answers multiple questions, exploring an architecture that manages a large number of applications with Argo CD (+10,000 apps, +250 clusters, etc.). This presents a challenge and is interesting for those exploring scalable environments.
• The article concludes that handling large clusters is hard for Argo CD, and we should consider HPA or splitting Argo CD into smaller components using tools like Kargo. To explore high availability (HA) provisioning, you can consult Argo CD HA setup and the benchmarking report from CNOE (Engineers from Intuit and AWS).

Security

1. Medium - SOC Analyst Automation: Scripts Every Analyst Should Be Using

• Every organization must prioritize security, especially their Security Operations Center (SOC), given the increasingly aggressive cybersecurity landscape. The author highlights the essential functions of a SOC and emphasizes the significant role of scripting in large systems, underscoring that automation consistently outperforms manual processes.
• You can delve deeply into this article to discover various solutions for facing attacks, including practical attachments, relevant tools, and multiple keywords to guide further exploration.

2. Medium - Secure APIs Without JWT: Alternatives You Should Know About 🌟 (Recommended)

• In this article, the author addresses the challenges and confusion surrounding the use of JSON Web Tokens (JWTs), highlighting instances where they are employed without full comprehension of their underlying mechanisms or inherent weaknesses. To mitigate these issues, the article provides a list of various alternative options to replace JWTs. Some of these techniques are genuinely new, highly interesting, and quite compelling.
• Each technology discussed is accompanied by examples, codebase snippets, and personal commentary on its advantages. This makes the article particularly helpful for readers seeking to choose the most suitable option for their specific needs. The technologies covered include PASETO, mTLS, HAWK, Session-Based Authentication, and Macaroons, among others.

3. Medium - AI Agentic Cybersecurity Tools: Reaper, TARS, Fabric Agent Action, and Floki

• AI is becoming increasingly relevant and ubiquitous in our daily lives, and the field of security is no exception. Numerous open-source projects are leveraging AI, with AI Agents being a particularly hot topic in recent days. Let’s delve deeply into this article to understand four notable AI Agents (Reaper, TARS, Fabric Agent Action, and Floki)
• This article offers an awesome walkthrough for anyone working in DevSecOps or SecOps, or for those interested in leveraging Large Language Models (LLMs) for exploitation. Through these tools, I see immense potential and exciting developments in this field.