The model never runs code. π§
Function calling in four steps, with validation before execution.
Day 23 of #30DaysOfAI. Full lesson π
#FunctionCalling#ToolUse
A chatbot answers and waits. An agent takes a goal and gets it done. π€
Perceive, plan, act, reflect. And always set a step limit.
Day 22 of #30DaysOfAI. Full lesson π
#AIAgents#AgenticAI
AI traffic breaks ordinary web networking. π
Streaming ingress, eBPF routing and default-deny egress.
Day 21 of #30DaysOfAI. Full lesson π
#KubernetesNetworking#eBPF
Your pod restarts and every file on its disk is gone. πΎ
Persistent volumes, access modes and StatefulSets keep AI state alive.
Day 20 of #30DaysOfAI. Full lesson π
#KubernetesStorage#VectorDatabase
Docker Compose runs your AI on one machine. What if it dies? βΈοΈ
Kubernetes: Pods, Deployments and Services for model serving.
Day 19 of #30DaysOfAI. Full lesson π
#Kubernetes#MLOps
Why is your AI container image 10 GB? π³
Multi-stage builds, cache mounts and a non-root user fix it.
Day 17 of #30DaysOfAI. Full lesson π
#Docker#DevOps
What if your AI never sent your data outside your network? π
Ollama for laptops. vLLM for GPU servers. Same OpenAI-style API.
Day 16 of #30DaysOfAI. Full lesson π
#LocalAI#Ollama#vLLM
Your RAG answer sounded confident. Was it right? π‘οΈ
Measure context relevance and faithfulness, and add an LLM judge that blocks unsupported answers.
Day 14 of #30DaysOfAI. Full lesson π
#RAG#AIart
https://t.co/lORabKZb62
What does it take to run an AI model on your own hardware? π₯οΈ
VRAM, optimized runtimes, and a headless Linux host.
Day 15 of #30DaysOfAI. Full lesson π
#AIInfrastructure#AI
https://t.co/l28aKAZkVE