AI & Machine Learning
Practical guides on AI infrastructure, agentic systems, LLM operations, and machine learning — built on open-source tools, sovereign infrastructure, and production engineering.
Featured Topics
LLM Observability & Eval-Driven Development — How to trace, evaluate, and monitor LLM applications in production: Langfuse, CI/CD eval gates, prompt versioning, drift detection, and cost-aware evaluation.
Agent Skill Management — How AI agents discover, manage, and execute skills across the 2026 ecosystem: MCP tools, OpenAI function calling, LangChain Deep Agents, CrewAI capabilities, the ACP protocol, skill registries, versioning, and production governance.
Agentic AI in Practice — A comprehensive deep dive from fundamentals to production ecosystems: core agent architecture, protocols (MCP/ACP/A2A), framework landscape, and deployment strategies.
Agentic AI Libraries Compared — A comprehensive comparison of LangChain, CrewAI, AutoGen, Semantic Kernel, and other agentic AI frameworks for building multi-agent systems.
Agentic Development: Beyond the Playbook — What open source teaches us about building software with AI agents. An alternative to Microsoft's corporate methodology.
Atlas Engine: Sub-2-Minute Cold Start — Run three specialised LLMs on a single DGX Spark with Hugging Face TGI and vLLM.
LLM API Value for Money Under $20/Month — A practical cost-benefit analysis of budget-friendly LLM APIs for self-hosted and small-scale deployments.
Vibe Coding in Production — How AI-assisted development reshapes the way we build software, and how to keep quality under control.
Hybrid Cloud-Local Inference Architecture — Production patterns for tiered inference routing that reduces costs by 50-100x by routing 70-80% of queries to local models while improving latency and privacy.
Agentic AI Governance and Security in 2026 — Bounded autonomy, inter-agent security, jailbreak prevention, and EU AI Act compliance for production agent deployments.
The On-Device LLM Revolution — How 3B-30B parameter models are moving to edge devices, driven by NPU hardware, quantization advances, and the privacy mandate.