AI Safety & Security Researcher

Giskard AI SAS; Paris; Sur place; EUR 60 000 - 90 000; Plein temps

Avantages offerts par ce poste

Remote-friendly

Résumé du poste

Giskard AI SAS recherche un chercheur en IA axé sécurité pour concevoir et appliquer des méthodes de test robustes des agents IA destinés à un usage enterprise.

Le poste exige un PhD en ML/IA ou expérience équivalente, une connaissance des LLM et une pratique du red-teaming. Localisation France ou relocalisation possible, conditions hybrides et package equity important.

Qualifications

• PhD en machine learning / IA ou expérience équivalente.
• Expérience avec Generative AI / LLMs et AI red-teaming.
• Localisation: France ou mobilité possible.
• Compétences en Python et code prêt pour la production.
• Capacité à communiquer des résultats complexes aux clients.

Responsabilités

• Concevoir et implémenter des méthodologies pour tester, évaluer et red-teamer des agents IA pour des cas d’entreprise, intégrées dans notre plateforme.
• Identifier activement les vulnérabilités et modes de défaillance des systèmes basés sur LLM, du prompt injection à l’exfiltration de données.
• Traduire des concepts de recherche en code prêt pour la production, du prototypage au déploiement.
• Réaliser une veille scientifique sur les domaines AI Testing / Benchmarking / Security.
• Présenter et documenter les résultats auprès des clients et partenaires (Mistral AI, AWS, Google, Microsoft, NVIDIA).
• Fournir expertise technique et recommandations de sécurité sur les projets clients.

Connaissances

LLMs

Cybersecurity

Applied research

Python

Git

English

Red team

Formation

PhD in ML/AI

Description du poste

Giskard defends the vision of a Responsible AI that serves the economic performance of companies while respecting the rights of citizens. To embody this mission, we are building a holistic testing platform for AI professionals to evaluate the quality & security of Generative AI applications based on Large Language Models (LLMs).

Since 2021, we’ve been building based on open-source foundations (4K stars on GitHub) and pushing the frontier of applied AI Safety & Security research, to help major customers secure their GenAI applications, including BPCE, SG, AXA, Michelin, and more.

Company track record

In 2022, we raised our first round of 1.5 million euros, led by Elaia, with notable angel investors including Hugging Face’s CTO and Mistral AI’s co‑founder. To read more about this fundraising, you can read our announcement.

In 2023, we received a strategic investment from the European Commission to build a SaaS platform to automate compliance with the EU AI Act. You can read more here.

In 2024, we won a 3 M€ grant from Bpifrance to lead research in evaluation & mitigation for LLMs, in a collaborative consortium with Mistral AI, Artefact, INA and BnF. You can learn more here.

Missions

Working directly with Giskard’s Lead AI Researcher, your missions will be to:

• Design and implement robust methodologies to test, evaluate, and red‑team AI agents for enterprise use cases, integrating them into our core platform.
• Proactively identify vulnerabilities and failure modes in LLM‑based systems, from prompt injection to data exfiltration to agentic misuse.
• Work closely with our engineering team to translate research concepts into production‑ready code, participating in the full development lifecycle from prototyping to deployment.
• Conduct our scientific watch, reading academic papers in the fields of AI Testing / Benchmarking / Security.
• Communicate findings with other AI teams at customers & partners, such as Mistral AI, AWS, Google, Microsoft, NVIDIA, in the form of presentations and papers.
• Provide technical expertise and actionable security recommendations on customer projects.

Requirements

We are looking for a researcher with an entrepreneurial mindset and a Ph.D. in Machine Learning / Artificial Intelligence or equivalent experience, preferably with previous experience with Generative AI / LLMs and AI application red‑teaming.

We are hiring people located in France or willing to relocate.

Skills

• Strong knowledge of Machine Learning / Artificial Intelligence research, with particular focus on LLMs and their failure modes.
• Solid cybersecurity fundamentals: threat modeling, common attack classes (injection, exfiltration, supply chain), and risk assessment, applied to AI/ML systems specifically.
• Demonstrated applied research experience with a focus on practical applications and tool development (papers, benchmarks, disclosed vulnerabilities).
• Strong skills in software engineering (use of Git, production‑ready code) and the Python stack. Comfortable steering and critically reviewing AI‑assisted workflows (coding agents, research assistant).
• Strong written & oral communication skills in English.
• Bonus: red‑teaming AI application or CTF experience, published security research, or open‑source contributions.

About you

• Passionate about building robust solutions to real‑life problems
• Comfortable moving between rigorous research and shipping something real that customers use
• Hungry for learning new skills every day
• Conscious about technology’s role in our society
• Empathetic team player

Benefits

We value diversity, equity, and inclusivity. Giskard is an equal opportunity employer. We do not discriminate on the basis of race, religion, color, origin, gender, sexual orientation, age, marital status, veteran, or disability status.

We are remote‑flex. We value productivity and work/life balance. You can come to our office in Paris on a weekly basis if you live around Paris. You can also work remotely and come to the office on a monthly basis.

We offer a remote office kit to all full‑time employees: a MacBook computer, external monitor, keyboard, mouse, etc. We also offer a very complete health insurance plan with Alan.

All full‑time employees are granted a significant equity package. We want to include everyone in the adventure and make it a collective win.

Salary: between 60 K€ and 90 K€ / year (brut) + 0.1 - 0.3% of equity