Courses
Three focused programmes for regulated organisations. Each starts with a free one-hour meeting to understand your environment and adapt the content. All available in English; questions and discussion in French.
AI Agents for Regulated Teams
Compliance officers, AML analysts, lawyers, legal ops, risk teams
Participants build a working AI agent during the session, adapted to their sector, workflows, and regulatory environment. The working example — AML alert, document review, fund reporting, or contract analysis — is chosen in the free one-hour meeting beforehand. Every output cites its source, every step is logged, and no result is acted on before a human has reviewed it.
- A working agent built and deployed during the session
- Every output traceable to its source document
- Human checkpoint wired into every workflow
- Certificate and session record for Article 4 compliance
AI in Practice: Supervised Build Workshop
Teams who have completed the flagship course and have a specific tool to build
You bring a specific workflow or problem from your environment. We build the tool together during the session under trainer supervision, with verification-first principles applied throughout. You leave with a tested, deployed solution your team owns completely.
- A working tool built around your specific workflow
- Full documentation for independent maintenance
- No vendor dependency: runs in your own environment
- Session record suitable for internal audit
AI Governance for Senior Management
CEOs, CCOs, Heads of Legal, Managing Partners, Chief Risk Officers
No hands-on building. Designed for leaders who need to evaluate what their teams, employees, and subcontractors produce with AI — and hold them to a standard. You leave knowing what compliant AI output must contain, what to demand, and what to reject.
- Ability to evaluate any AI tool your teams or subcontractors bring you
- Framework for what compliant output must contain
- EU AI Act deployer obligations in plain language
- Confidence to answer your regulator, board, or auditors
All courses are delivered in-house, at your premises, for your team. House of Training and Digital Learning Hub, Luxembourg, host public versions of these courses. Pricing and scheduling on request.
In Action
An AML transaction monitoring alert — the kind your compliance team handles daily. This is what the training produces: an agent your team built, in their own environment, where every conclusion traces back to a source they can check before anyone signs off.
Six cash deposits over 9 days: EUR 14,800 / 14,850 / 14,900 / 14,950 / 14,800 / 14,950. Total: EUR 89,250. Each deposit individually below the EUR 15,000 mandatory identification threshold.
Compliance officer task: assess the pattern, retrieve the applicable regulation, and produce a documented memo for officer sign-off before any filing decision.
Your compliance team builds this agent themselves, in your environment, with traceability by design.
I remain available as patterns grow more complex and the regulatory environment evolves.
Pattern: Six deposits EUR 14,800 to EUR 14,950 over nine days. Total EUR 89,250. Pattern consistent with fractionnement — deliberate structuring below the identification threshold.
Assessment: Grounds for a suspicious transaction report to the CRF.
- Pattern identified as fractionnement → Loi du 12 novembre 2004
- EUR 15,000 threshold confirmed → Loi du 12 novembre 2004
- STR obligation triggered → Loi du 12 novembre 2004
- Audit trail logged → CSSF AML/CFT requirements
- No action taken pending human sign-off
Every conclusion cites its source. Every step is logged. Nothing moves until a human signs off. That is traceability built in — not described in a policy, but wired into the agent your team builds during the course.
This is the assessment memo the agent produces — the kind your compliance team builds during the Finance course. Every regulatory citation is generated by the agent and logged before the officer sees it.
Sample output from the Finance course. All regulatory citations are generated and logged by the agent before the compliance officer reviews.
How I Work
Every engagement starts with a free conversation to understand your environment. No generic presentation.
One hour to understand your context
Before any training or engagement begins, I spend one hour with you to understand your organisation's environment, tools, workflows, and regulatory context. The course content and working examples are adapted accordingly. This meeting costs nothing and commits you to nothing.
Training: your team leaves with a working agent, not just knowledge
I train your teams to get the most out of the AI tools they already have, starting with the standard chatbots available in your environment (Mistral Le Chat, Copilot, or similar). Every session is built around full GDPR compliance and the four obligations of the EU AI Act: literacy, transparency, human oversight, and traceability. By the end, your team does not just use AI; they use it responsibly, and you have documented proof of compliance with Article 4.
Fractional AI Officer: ongoing expert presence
For teams that need continuous, adaptive support after the training. I act as your fractional AI officer, available according to your pace and your needs, keeping your AI use current and compliant as your operations evolve and as the regulatory environment develops. No full-time hire needed.
Transition Management: direct leadership of the transformation
For teams that need someone to step in and lead the change end to end. I come in for a defined period, typically one to two months, take direct responsibility for the AI transition, build the first compliant pipelines with your team, and hand over a fully operational, AI Act-compliant setup. Two fee options: fixed price, or fixed costs plus a share of the savings I generate.
About
Your teams are using AI every day. The question your board, your auditors, and your regulator will ask is not whether they use it — it is whether they use it in a way they can explain, document, and stand behind.
Can they trace every output back to its source? Is there a human decision point before anything is acted on? Do they know when to trust the result and when to stop? Can they show an auditor what happened and why? Can they explain a decision to a client or a regulator without opening a black box? Article 4 on AI literacy is already in force. High-risk obligations follow by December 2027. My training answers all of them — not by teaching rules, but by having your teams build agents where traceability is baked in by design: every output cites its source, every step is logged, and nothing reaches a human decision-maker without a clear chain of evidence they can read and verify.
Why It Stays Current
As a peer reviewer for leading international AI conferences and journals, an evaluator for EU Horizon and Marie Sklodowska-Curie Actions funding calls, and an examiner of PhD theses, I see the research before it is published and judge where Europe is putting its most serious research funding. I combine active research evaluation with production deployment: I review what is coming before it is public and ship systems this week. That is what I bring to every engagement.
Now
AI Trainer
2016 - present
American University of Paris (2016–2024) · Digital Learning Hub, House of Training, and public and private organisations, Luxembourg (2025–present)
- Tenured professor at the American University of Paris (2016–2024): designed and launched the university's first Data Science programme, modelled on UC Berkeley; taught 40+ courses across computer science, mathematics, statistics, and AI ethics
- Since 2025: AI literacy and adoption programmes for professional and public-sector teams, built around transparency, traceability and human oversight — satisfying EU AI Act Article 4 with documented outcomes
- Formats from half-day to full-day; governance modules for managers and legal referents
- 9.13 / 10 overall satisfaction rate · 4.0 / 4 trainer rating
AI Transition Manager and Fractional AI Officer
2025 - present
Independent, Luxembourg and beyond
- For teams that need ongoing expert guidance: fractional AI officer, adapting to the organisation's pace and needs
- For teams requiring full transformation support: direct transition management, typically one to two months, to solve a defined AI problem end to end
- Both models leave your team able to operate independently and in full compliance
Co-founder and CEO
2023 - present
Enidia AI, Luxembourg
- Verification-first, on-premise AI for regulated industries; GDPR and EU AI Act by design
- Clients include a top-10 international law firm
- Production document review cut from about 25 hours to about 4
Research and Academia
Assistant Professor (Tenured)
2016 - 2024
The American University of Paris, France
- Designed and launched the university's first Data Science programme, modelled on UC Berkeley
- Taught 40+ courses across computer science, mathematics, statistics, and AI ethics
Principal Investigator
2020 - 2025
University of Luxembourg, Luxembourg
- Principal investigator of more than 600,000 euros in public funding, including an FNR CORE grant
- Led the access-to-justice line: AI that helps lay users navigate the law. Patent law assistant demo
Postdoctoral Researcher
2012 - 2016
Microsoft Research and Inria Joint Centre, Paris
- Software verification and proof assistants, in the teams of Leslie Lamport and Dale Miller
Earlier Industry
Programming Team Lead
2001 - 2006
Quigo Technologies, New York, USA
- R and D lead at an AI-driven advertising startup; the company was acquired by AOL for about 360 million dollars
Education
Ph.D., Computer Science
2008 - 2012
Technical University of Vienna, Austria
M.Sc., Computer Science
2006 - 2008
Technical University of Vienna, Austria
B.Sc., Mathematics and Computer Science
1998 - 2001
Hebrew University of Jerusalem, Israel
Standing
33 international refereed conference papers and 3 journal papers. 402 citations, h-index 12. Reviewer for international journals and for Horizon and Marie Sklodowska-Curie Actions, with a best-paper award among co-authored work. Creator of the open-source GAPT proof system.
Built to Be Trusted
Standard AI tools produce outputs with no indication of whether they are correct. In regulated environments, an unverified output is a liability. Every system I build and every course I deliver follows a verification-first principle: every output cites its source, every step is logged, and nothing reaches a decision-maker without a chain of evidence they can read and check.
at a top-10 international law firm
Bespoke AI Solutions
Training is where teams learn to use AI. Enidia AI is where we build it for them — verification-first, on-premise, for regulated sectors.
Bespoke AI Pipelines
Built specifically for your environment, your documents, and your workflows. On-premise, on open-weight models. No vendor dependency, no token costs, no documents leaving your infrastructure. Designed to satisfy CSSF, AML, and EU AI Act requirements from the ground up.
Migration from Cloud AI Tools
Teams using Harvey, Leora, Copilot, or similar cloud tools face token costs, data residency risk, and vendor dependency. We migrate these workflows to on-premise open-weight models — keeping the capability, removing the exposure. Your team keeps what they built; the documents stay inside.
at a top-10 international law firm
Verification-first pipeline built by Enidia AI
Media Center
Selected media appearances and speaking engagements.
- All
- Programs
- Items
- Web
Testimonials
In their own words, from learners at the Digital Learning Hub.
Contact
Want to build something, de-risk an AI you already have, or get your team using AI well? Email tomer [at] libal.info or connect on LinkedIn.
Powered by