Medium

Miodrag Cekikj

In

AI Advances

by

Miodrag Cekikj

·

1d ago

Give the Knowledge Layer the Decision: An Agent Above Foundry IQ and the Fabric IQ Ontology…

The last article ended with the knowledge layer holding both of its sources and lacking one thing: the decision about which to ask, in what…

Give the Knowledge Layer the Decision: An Agent Above Foundry IQ and the Fabric IQ Ontology…

In

AI Advances

by

Miodrag Cekikj

·

Oct 26

Give the Knowledge Layer Something to Compute: Foundry IQ and the Fabric IQ Ontology, Measured

In one of my previous articles I measured Microsoft’s knowledge layer over documents and found a wall on the questions that need an…

Give the Knowledge Layer Something to Compute: Foundry IQ and the Fabric IQ Ontology, Measured

In

AI Advances

by

Miodrag Cekikj

·

Oct 23

Extract or Load? I Built the Same Knowledge Graph Both Ways and Measured the Difference

The first article in this series claimed that loading structure from systems of record is cheap and extracting it from documents is…

Extract or Load? I Built the Same Knowledge Graph Both Ways and Measured the Difference

In

AI Advances

by

Miodrag Cekikj

·

Oct 22

Microsoft’s Knowledge Layer, Measured

Every major platform now sells a knowledge layer for enterprise agents. I put Microsoft’s through the same questions, data and judge as my…

Microsoft’s Knowledge Layer, Measured

In

AI Advances

by

Miodrag Cekikj

·

Oct 21

What Breaks When the World Moves and Your Ontology Does Not

Every measurement in this series assumed the graph, the pages and the ontology agreed with the company. In a real company they drift apart…

What Breaks When the World Moves and Your Ontology Does Not

In

AI Advances

by

Miodrag Cekikj

·

Oct 20

How Much Meaning Does Your Graph Need?

An ontology is not a document. It is a budget. I cut the one behind my knowledge graph into six layers, gave the agent one layer at a time…

How Much Meaning Does Your Graph Need?

In

AI Advances

by

Miodrag Cekikj

·

Oct 14

Not a Retrieval Problem. A Shape Problem.

Better retrieval finds better passages. It cannot find what is between them. Three designs on one corpus, measured by the shape of the…

Not a Retrieval Problem. A Shape Problem.

In

AI Advances

by

Miodrag Cekikj

·

Oct 14

Route First, Retrieve Second

The cheapest call in your AI stack is the one that decides the bill. I put a question router in front of three retrieval designs and…

Route First, Retrieve Second

In

AI Advances

by

Miodrag Cekikj

·

Oct 13

Written Down or Worked Out? The Two Questions That Decide Your AI Architecture

I took the rule of thumb I use to choose between RAG, a knowledge graph and an agent, and put it on trial: twenty questions, four kinds…

Written Down or Worked Out? The Two Questions That Decide Your AI Architecture

In

AI Advances

by

Miodrag Cekikj

·

Oct 6

Your Data Has a Shape. Does Your AI Know It?

I built Classic RAG, GraphRAG and an agentic hybrid on the same data and measured them on the same questions, with six models. The model…

Your Data Has a Shape. Does Your AI Know It?

Miodrag Cekikj

Miodrag Cekikj

Founder of Adopt Intelligence| PhD Computer Science & Engineering | Microsoft MVP | Technical Trainer | AI/ML Practitioner | Specialized in AI Agents & Graphs

Following

Help

Status

About

Careers

Press

Blog

Store

Privacy

Rules

Terms

Text to speech