Fractional Chief AI Officer

I find what your AI
doesn't know it's missing.

Most AI systems tell you what's in the data. I've built systems that tell you what's conspicuously absent — and proven those predictions against ground truth.

This isn't prompt engineering with a knowledge graph wrapper. It's graph ML research — temporal validation, ComplEx embeddings, contrastive learning — applied to your domain.

Book a CAIO Discovery CallView void-graph on GitHub

Validated Results

Not a demo. Not a prototype. Temporal validation against real-world data that emerged post-cutoff.

0.94
Precision@100
Biomedical
Hetionet + STRING · 47K nodes · 2.25M edges
0.86
Precision@100
Cybersecurity
MITRE ATT&CK knowledge graph

What is Anti-GraphRAG?

01
Freeze the graph
Construct a knowledge graph frozen at time T. Everything that exists is captured. Everything that should exist but doesn't is the target.
02
Train void embeddings
Train ComplEx embeddings on the existing graph. Apply a Contrastive Void Network — a contrastive head trained to distinguish meaningful absences from random gaps.
03
Validate against reality
Score absent edges. Validate top predictions against what actually appeared in the real world after the cutoff date. Precision@100 is the metric.

Who This Is For

Industries where meaningful absence — not just presence — is the competitive edge.

Pharma & Biotech

Detect missing drug-target interactions, undocumented compound-pathway links, and competitor pipeline gaps before they cost you.

Cybersecurity

Surface MITRE ATT&CK coverage gaps, predict threat actor technique expansions, and validate mitigation completeness.

Legal & RegTech

Find unlinked compliance obligations across EUR-Lex, GDPR, CFR, and FIBO knowledge graphs before regulators do.

Financial Intelligence

Identify missing connections in market knowledge graphs — relationships between entities that should exist but don't yet.

What You Get

Knowledge Graph Audit

Map your existing knowledge infrastructure, identify structural gaps, and assess what your current AI systems are blind to.

Void Detection Pipeline

Build and deploy the Anti-GraphRAG methodology on your domain data — trained, validated, and production-ready.

AI Strategy & Roadmap

Define where AI creates real leverage in your business, what to build vs. buy, and how to measure whether it's actually working.

Team Enablement

Upskill your existing data/ML team to own and extend the knowledge graph infrastructure independently.

Engagement Models

Strategic Advisory

1 day/week
3–6 months

You need a sounding board and strategic direction. Your team does the building.

Most Common

Build + Advisory

2–3 days/week
3–6 months

You need both strategic ownership and hands-on knowledge graph engineering.

Proof of Concept

Fixed scope
4–8 weeks

You want a working Anti-GraphRAG pipeline on your domain data before committing to ongoing work.

Ready to find what's missing?

30-minute discovery call. We'll identify whether your domain has meaningful absence worth surfacing and whether Anti-GraphRAG is the right tool for it.

Book a Discovery Call