AI Security Researcher

Job Description:

As an AI Security Researcher focused on LLM and agent red teaming, you will further build and enhance our automated red-team agent. You will research new adversarial methods, design evaluation frameworks, and develop systems that discover vulnerabilities in LLM & multimodal agents.

This is an applied research role with strong product impact. You will move from open-ended research questions to working prototypes, evaluation pipelines, and production capabilities in the DeepKeep platform.

About DeepKeep:

At DeepKeep, we create a comprehensive solution for safeguarding and evaluating ML applications in multiple fields, including Large Language Models, Computer Vision, and multi-modal applications.

The DeepKeep Platform protects models in real-time from adversarial attacks, problematic outputs and data leakage, while also enabling a wide range of evaluation methods to measure model performance.

We take pride in being on the forefront of technology and always strive to develop and implement the best solutions there are.

Key Responsibilities:

• Research and implement new attack methods, including prompt injection, jailbreaking, tool misuse, RAG attacks, context/memory poisoning, data leakage, and model manipulation.
• Design adversarial scenarios against tool-using agents and AI applications in authorized evaluation environments.
• Build evaluation harnesses, red-team benchmarks, scoring methods, and regression tests.
• Analyze model and agent failures across multi-step reasoning, tool use, memory, retrieval, and workflow orchestration.
• Fine-tune, distill, or adapt small language models when useful for attack generation, vulnerability classification, or evaluation.
• Translate research into product capabilities, mitigations, technical guidance, and customer-facing value.
• Collaborate with researchers, engineers, and product teams to shape DeepKeep’s AI security roadmap.

Qualifications:

Professional Experience:

• MSc. / PhD in Computer Science, Mathematics, or a related data-focused field And 2 years of data science / security research experience.
• Or BSc. in Computer Science, Mathematics, or a related data-focused field And 4 years of data science / security research experience.
• Agents & tool-using LLMs: Experience building, evaluating or red-teaming agentic systems.
• Cyber Security & Red Teaming: Experience in identifying, exploiting, and/or mitigating security vulnerabilities through hands-on research & adversarial testing.

Professional Experience Advantages:

• NLP Methodologies: Experience in NLP-specific tasks, like training transformer-based models etc.
• Software Engineering: Experience in software development in production environments.
• Research Experience in conducting theoretical or experimental research, using statistical tools and mathematical formulations to solve complex problems.
• Technical Skills: Proficiency in Python.
• Excellent interpersonal skills for effective collaboration and teamwork.