Current Projet:

In the Turbo Team, designing a neuro-symbolic constraint solver, GPU compliant and supporting abstract domains. My current project focuses on opening Turbo to constraint acquisition techniques, tailoring LLMs for breaking down the expert barrier of Constraint Programming technologies.

Previous Projets:

𝐔𝐧𝐢𝐟𝐲𝐢𝐧𝐠 𝐚𝐧𝐝 𝐞𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐭 𝐈𝐑 𝐫𝐞𝐩𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐭𝐫𝐚𝐧𝐬𝐜𝐨𝐦𝐩𝐢𝐥𝐢𝐧𝐠 𝐚𝐧𝐝 𝐪𝐮𝐚𝐥𝐢𝐭𝐲 𝐭𝐞𝐬𝐭 𝐬𝐮𝐢𝐭𝐞 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧.

MTB111

As a researcher in the R&D team, I contributed to designing and developing Structur’All X, a JAVA framework to transcompile and generate quality test suites based on a generic intermediate representation within a graph database (OrientDB & Gremlin).

---

𝐓𝐚𝐢𝐥𝐨𝐫𝐢𝐧𝐠 𝐇𝐲𝐛𝐫𝐢𝐝 𝐅𝐮𝐳𝐳𝐢𝐧𝐠 𝐟𝐨𝐫 𝐝𝐞𝐭𝐞𝐜𝐭𝐢𝐧𝐠 𝐕𝐮𝐥𝐧𝐞𝐫𝐚𝐛𝐢𝐥𝐢𝐭𝐢𝐞𝐬 𝐢𝐧 𝐈𝐎𝐓 𝐍𝐞𝐭𝐰𝐨𝐫𝐤 𝐒𝐭𝐚𝐜𝐤𝐬.

Uppsala University

In the aSSIsT project, I designed an automatic framework to statistically evaluate hybrid fuzzers (QSYM, Angora, SymCC) and sanitisers (ASan, UBSan, EffectiveSan). Based on Docker / Perl scripts, the framework was applied to Contiki-NG’s network stack, exposing several vulnerabilities.

---

𝐈𝐦𝐩𝐥𝐞𝐦𝐞𝐧𝐭𝐢𝐧𝐠 𝐚𝐧 𝐀𝐛𝐬𝐭𝐫𝐚𝐜𝐭 𝐚𝐧𝐝 𝐒𝐲𝐦𝐛𝐨𝐥𝐢𝐜 𝐄𝐱𝐞𝐜𝐮𝐭𝐢𝐨𝐧 𝐨𝐧 𝐚 𝐓𝐢𝐧𝐲 𝐈𝐧𝐭𝐞𝐫𝐩𝐫𝐞𝐭𝐞𝐫.

Salzburg University

Developped Littlemonster, an extention Selfie’s interpreter (written in Cstar and interpreting RISC-V) with symbolic execution based on abstract domains.

---

𝐏𝐡𝐃: 𝐌𝐨𝐝𝐞𝐥-𝐁𝐚𝐬𝐞𝐝 𝐓𝐞𝐬𝐭𝐢𝐧𝐠 𝐨𝐟 𝐈𝐧𝐭𝐞𝐫𝐚𝐜𝐭𝐢𝐯𝐞 𝐌𝐮𝐬𝐢𝐜 𝐒𝐲𝐬𝐭𝐞𝐦𝐬.

IRCAM and Sorbonne University

I designed automatic methods based on Timed Automata to establish the timed conformance of Antescofo (C++).

(thesis)

---

—