https://mkadri85.github.io/blog/ Practical, tool-agnostic guides on AI-assisted engineering and running real production systems with small teams. en Wed, 23 Sep 2026 14:00:00 +0000 https://mkadri85.github.io/blog/five-thousand-alarms-one-ticket/https://mkadri85.github.io/blog/five-thousand-alarms-one-ticket/Wed, 23 Sep 2026 14:00:00 +0000One fiber cut, five thousand alarms, one ticket: alarm correlation and root cause analysis stage by stage, demarcation, and the arrival-order problem. https://mkadri85.github.io/blog/topology-restoration-alarm-correlation/https://mkadri85.github.io/blog/topology-restoration-alarm-correlation/Tue, 22 Sep 2026 00:00:00 +0000How a multi-vendor assurance platform reconstructs network topology from raw exports, why alarm correlation depends on that map, and the five places it breaks. https://mkadri85.github.io/blog/self-host-llm-one-gpu/https://mkadri85.github.io/blog/self-host-llm-one-gpu/Mon, 17 Aug 2026 00:00:00 +0000One week, one cloud GPU, a shelf of open models - and the discovery that the model is only one layer of an engineered system. Quantization, the 32B wall, the harness lesson, guardrails, and real GCP prices. https://mkadri85.github.io/blog/isac-cell-towers-as-sensors/https://mkadri85.github.io/blog/isac-cell-towers-as-sensors/Mon, 03 Aug 2026 00:00:00 +0000Cell towers as sensors: what 5G ISAC really does today, how big the wireless foundation models actually are, and the sensing business nobody has built. https://mkadri85.github.io/blog/mixture-of-experts-explained/https://mkadri85.github.io/blog/mixture-of-experts-explained/Mon, 27 Jul 2026 00:00:00 +0000A zero-background guide to Mixture of Experts: experts, routers, total vs active parameters, and why a 2.8 trillion parameter model activates only 104 billion per word. https://mkadri85.github.io/blog/why-long-context-is-hard/https://mkadri85.github.io/blog/why-long-context-is-hard/Mon, 27 Jul 2026 00:00:00 +0000Part 2 of a zero-background series: why the memory of a long conversation is the hidden cost of AI, and the fixes - GQA, MLA and KDA - explained simply. https://mkadri85.github.io/blog/agentic-ai-microwave-backhaul/https://mkadri85.github.io/blog/agentic-ai-microwave-backhaul/Fri, 17 Jul 2026 00:00:00 +0000What an agentic AI closed loop must actually reason about on microwave backhaul: rain fade vs hardware ambiguity, multi-vendor OSS islands, blast-radius-gated actions, and verification - with an open live demo. https://mkadri85.github.io/blog/agent-noc/https://mkadri85.github.io/blog/agent-noc/Mon, 13 Jul 2026 00:00:00 +0000What thirty years of telecom network-operations discipline transfers to AI agent fleets, what breaks in translation, and the assurance-tooling gap nobody has closed. https://mkadri85.github.io/blog/agentic-sre-agent-fleets/https://mkadri85.github.io/blog/agentic-sre-agent-fleets/Thu, 09 Jul 2026 00:00:00 +0000A field guide to applying the SRE incident loop to autonomous AI agents themselves: detection, a control plane, and a circuit breaker that keeps autonomy from making incidents worse. https://mkadri85.github.io/blog/shared-ai-memory/https://mkadri85.github.io/blog/shared-ai-memory/Mon, 06 Jul 2026 00:00:00 +0000A step-by-step, tool-agnostic guide to giving every developer and every AI coding agent on a team the same project memory.