Gen-AI Developer Classroom notes 26/May/2026

Model (LLM)

• A model at its core predicts the next token
• A Large Language Model is trained on almost all the publically available web information, books, research papers, opensource code
• A model can generate different types of outputs (Modal)
  • Text
  • Image
  • Video
  • Code
• A multi modality LLM

What we want

• I want an LLM which looks intelligent, to

  • Create a leave letter and send to my teacher (LLM + email sending code)
  • Create a Motivational quote for family whatsapp group + Send to whats app group

• Initial phase:

  • Specific code for actions (we call llm programatically + perform action.)
• LLM’s introduced tool calling
  • This helped in llm deciding which tool to call.
• Prompt

Simulate the tool calling outputs, Assume i have tools add, sub, div, mul, mod and now i have asked llm the following question " i have 5 chocolates purchased 3 more , How many chocolates will i have"

• Prompt

Simulate the tool calling outputs, Assume i have tools 
start server, stop server, restart server, delete server and now i have asked llm the following question " The server host-prd-1 with ip 10.100.10.11" is not responding, the cpu is showing 100% solve this

• ReACT Agent: Reason + Act
  

Problems

• LLMs work with a principle Garbage in => Garbage out i.e. model output will be good according to your prompt

• Models have context limits => number of tokens that can be passed and generally if you look at llm pricing we started

  • Context Engineering.
  • Problem to solve: Context Poisoning.

• Gaurdrails

• Use Grounding to solve Hallucinations