Causal AI Research & Consulting
Systems that explain why, not just predict what.
NeuralKer is David Granados' research and consulting practice on causal inference for AI systems — from physically-verified autonomous vehicle safety benchmarks to causal retrieval-augmented generation. Substance before hype.
The approach
Most AI systems stop at correlation. This work climbs Pearl's Ladder of Causation.
Click a rung — the diagram on the right shows what changes at each level, using the same example throughout: a car approaching a hazard.
Selected publications
Verified with physics, not assumptions.
Only flagship papers and long-form pieces live here, curated — not a mirror of every weekly LinkedIn post or Medium article. Only the most recent gets the full write-up below; earlier work compacts into a single row, so this page can't grow forever.
About
David Granados
PhD candidate in the Doctoral Program in Automatic Control at the Universitat Politècnica de Catalunya (UPC), researching causal inference for safety-critical AI. NeuralKer is where that research meets applied consulting — for teams whose AI systems need to survive an audit, not just a demo.
- FocusCausal inference, structured knowledge systems, AI safety verification
- ProgramUPC Doctoral Program in Automatic Control
- Based inBarcelona, Spain
Who this is for
Regulated industries
Pharma, legal, finance — where a model's reasoning has to survive an external audit, not just a demo.
Structured knowledge bases
Teams building RAG systems where every answer needs to trace back to a real, checkable source.
Safety-critical pipelines
Systems where "it usually works" isn't good enough, and where the data itself needs auditing first.
Get in touch