Loading...
DevOps & Cloud Engineering Journal
Insights for Builders
Deep dives into Kubernetes, CI/CD, and Cloud Architecture. Practical guides for modern infrastructure teams.
⌘K
AI & Data
9 min readLocal LLM VRAM Requirements: The Math Behind Which Models You Can Run
Every 'best local model' list goes stale in months. The arithmetic underneath it doesn't. Here's how to compute the memory a model needs from its parameter count and quantisation level, why the KV cache is the number people forget, and how mixture-of-experts broke the old rule that parameters predict speed.
Ajeet YadavJuly 28, 2026
Read Article Latest Articles
272 posts
AWS
9 min readEKS Auto Mode for GPU Workloads
Auto Mode ships NVIDIA drivers and the device plugin, so the GPU Operator stops being your problem. In exchange you get a 14-day node expiry, weekly AMI replacement, and no SSH — which matters a lot more for training than for inference.
By Coding Protocols TeamRead
AI & Data
12 min readScale-to-Zero LLM Inference on Kubernetes
An idle GPU pod bills exactly the same as a saturated one. Here is how to scale LLM inference to zero replicas with KEDA, and how to cut the cold start from eight minutes to under two.
By Coding Protocols TeamRead
AI & Data
10 min readTrack LLM Inference Cost in Kubernetes With OpenCost
Your GPU node bill is easy to find. Your cost per million tokens is not — and it's the only number that compares to API pricing. Here's how to derive it by joining OpenCost's dollars to vLLM's token counters.
By Coding Protocols TeamRead
Security
11 min readLeast-Privilege IAM for AI Agents
ReadOnlyAccess grants ssm:Get*, s3:GetObject and lambda:GetFunction — it is a credential-read policy wearing a safe-sounding name. Here's how to scope an AWS role for an agent that inspects infrastructure.
By Coding Protocols TeamRead
Platform Engineering
12 min readHuman Approval Workflows for AI Infrastructure Agents
A gate that fires on every action gets rubber-stamped within a week, and then it's worse than no gate. Tier by reversibility, show the approver enough to judge, and re-validate at execution.
By Coding Protocols TeamRead
Observability
9 min readTrace MCP Tool Calls With OpenTelemetry
Your agent trace stops at the MCP boundary. HTTP headers can't fix it — stdio has none, and one Streamable HTTP connection carries many tool calls. The fix is the protocol's own _meta field.
By Coding Protocols TeamRead