← back to inferent-it.com
Lessons · long-form technical writing

How I actually think
through hard systems problems.

Deep dives into agentic AI architecture — LangGraph, RAG, multi-agent frameworks — written the way I'd explain them on a whiteboard. No fluff, no code in the reasoning, just the judgment calls and why the obvious shortcut usually fails.

01
RAG series · architecture
Architecting Advanced RAG — From Zero to Deployment
How to think through a production-ready, domain-agnostic RAG system before writing a line of code — the retrieval, grading, and retry decisions that actually matter.
LangGraphRAGno code
Read the lesson →
02
Foundations
How Does an LLM Actually Learn?
From a blank model that knows nothing to a model you can hold a conversation with — weights, backpropagation, pre-training, and RLHF, with interactive visuals.
LLM basicsinteractive
Read the lesson →
03
Coming soon
Multi-Agent Frameworks, Compared on Their Own Terms
LangGraph, CrewAI, BeeAI, and AG2 — what each one actually assumes about orchestration, and where those assumptions break.
CrewAIBeeAIAG2
In progress
04
Coming soon
Reflection, Reflexion, and Learning Agents
How agents get better without retraining — self-critique loops, memory, and the difference between reacting and actually learning.
agentsreflection
In progress
05
Coming soon
State Is Everything
Why treating the LLM as a stateless function called by a stateful machine is the single idea that makes agentic systems predictable.
state designarchitecture
In progress