Skip to content
View Yarmoluk's full-sized avatar
💭
At Peace
💭
At Peace

Highlights

  • Pro

Block or report Yarmoluk

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Yarmoluk/README.md

Daniel Yarmoluk

Building pre-structured knowledge graphs that outperform RAG at 11× lower token cost.

Founder of Graphify.md — CKG architecture, benchmark design, and enterprise deployment.


Compact Knowledge Graph Benchmark

42× more intelligence per token than RAG. Zero hallucinations by construction.

45 domains · 7,928 queries · Fully reproducible

System Macro F1 Tokens/query RDS
CKG 0.4709 269 0.00175
RAG 0.1231 2,982 0.0000413
GraphRAG 0.1200 3,450 0.0000452

CKG F1 improves with hop depth (0.374 → 0.772 at hop=5). RAG plateaus at hop=2.

ckg-benchmark — benchmark repo, paper, reproducible results


What Is CKG?

Pre-structured knowledge as a plain .md or .csv file. Drop it in your LLM context. No graph database, no embeddings, no retrieval pipeline.

The structure is the signal — not the curation effort. Track 2 proved this: a GLP-1/Obesity pharmacology CKG built entirely from the ClinicalTrials.gov API in one session achieved F1 = 0.530, exceeding the hand-curated educational average.


Active Projects

Project What it is
ckg-benchmark 45-domain RAG vs CKG vs GraphRAG benchmark — paper + data + code
ckg-mcp MCP server — CKG as a pre-compiled routing layer for agent stacks (coming)
Graphify.md Commercial CKG deployment — weekly-updated domain knowledge for enterprise AI

Paper

Benchmarking Knowledge Retrieval Architectures Across Educational and Commercial Domains: RAG, GraphRAG, and Compact Knowledge Graphs Yarmoluk & McCreary, 2026 · v0.6.2 · Patent pending App #64/040,804

Read the paper · PDF


Contact

graphifymd.com · LinkedIn

Popular repositories Loading

  1. ckg-benchmark ckg-benchmark Public

    Benchmarking RAG vs GraphRAG vs Compressed Knowledge Graphs on the McCreary Intelligent Textbook Corpus

    HTML 5

  2. Yarmoluk Yarmoluk Public

    GitHub Profile

  3. ckg-mcp ckg-mcp Public

    Compact Knowledge Graph MCP server — 42× more efficient than RAG on structural queries

    Python

  4. ckg-skills ckg-skills Public

    Claude Code skills for building Compact Knowledge Graphs — 42x more efficient than RAG

  5. skills-1 skills-1 Public

    Forked from anthropics/skills

    Public repository for Agent Skills

    Python