Benchmark design
Mathematical problems and grading standards for evaluating AI.
- Original problems and reference solutions
- Grading rubrics and error categories
- Checks for ambiguity and unintended shortcuts
Independent consulting · Strasbourg, France
Mathematical judgment for AI research.
I help research teams evaluate AI reasoning, design mathematical benchmarks, and review complex proofs.
Research mathematician · CNRS
PhD · HDR · 30+ invited talks
Research
judgment
judgment
What I do
Put the reasoning
to the test.
Original benchmarks for AI teams. Independent reviews of difficult arguments.
Mathematical review
Independent review of proofs and arguments from people or models.
- Assumptions, gaps, and edge cases
- Proof outlines for formalisation
- Written findings and recommendations
About
An eye for
hidden assumptions.
I am a CNRS research mathematician at IRMA in Strasbourg, working in arithmetic geometry. I previously held fellowships at IMJ–PRG in Paris and the Max Planck Institute for Mathematics in Bonn.
My consulting draws on the same habits as my research: question the assumptions, check each step, and make the conclusion precise.
At a glance
Based inStrasbourg, France
FocusAI evaluation · Mathematical review
LanguagesEnglish · French · Italian
AvailabilityProject-based
Independent consulting, subject to institutional authorisation and conflict checks.
How I work
Define the question.
Follow the evidence.
01
Set the scope
Agree on the question, the standard, and the deliverable.
02
Check each step
Test the assumptions, logic, and edge cases.
03
State the limits
Separate what is established from what remains uncertain.
04
Explain the findings
Present the conclusions and recommend the next steps.
Have a project in mind?
Let’s work through
the hard part.
Send me the question, the result you need, and your timeline. I’ll let you know how I can help.
daddezio.math@gmail.com