IT-AI-Base

From wiki.samerhijazi.net
Jump to navigation Jump to search

init

Buzz Words

  • LLM (Large Language Model)
  • LLMOps
  • Generative KI
  • RAG (Retrieval Augmented Generation): Ein LLM nutzt zusätzlich deine eigenen Daten, bevor es antwortet.
  • GenAI und Agentic AI Tools (z. B. LlamaIndex, LangChain/LangGraph).
  • LangGraph-orchestrated LLM agents to autonomously investigate cloud incidents
  • PagerDuty, Datadog, Grafana, Netdata, Dynatrace, Coroot, ThousandEyes, BigPanda
  • Any tool that can send webhooks can trigger an automated investigation.
  • How is Aurora different from Rootly or incident.io?
  • LangChain4j: is an open-source Java library designed to simplify integrating Large Language Models (LLMs) into JVM applications

ML-Frameworks

  • TensorFlow, PyTorch oder scikit-learn

LLM-Modell

  • Anthropic (Claude: Opus 4.7, Sonnet 4.6, Haiku 4.5)
  • OpenAI → (GPT Modelle)
  • Meta → (LLaMA)
  • Google → (Gemma / Gemini)
  • Microsoft → (Phi)
  • Mistral AI → Mistral / Mixtral
  • Alibaba → (Qwen: Qwen 3.5 / 3.6-Plus)
  • DeepSeek → (deepseek-coder)
  • BigCode → (StarCoder)
  • Hugging Face → (Hosting + Modelle)

LLM

  • A Large Language Model (LLM) is an AI system trained on huge amounts of text to understand and generate human-like language.
  • An LLM is a program that predicts the next word in a sentence, based on everything it has learned from massive datasets.
  • LLM = very advanced autocomplete trained on the internet

1. What LLMs can do

  • Answer questions
  • Write emails, essays, code
  • Translate languages
  • Summarize documents
  • Chat like a human

2. What LLMs can't do

  • LLMs don’t “think” like humans.
  • Don’t truly understand meaning
  • Don’t have opinions
  • Just predict the most likely next words based on patterns

3. Examples of LLMs

Regulatory Requirements

  • Regulatory requirements (MaRisk, BAIT, DORA)
  • BAIT, DORA oder ISO 27001.