PhD in RL‑Driven LLMs for Secure Code Analysis

1000scholars; Rennes; On-site; EUR 24,000 - 30,000; Full time

Job summary

1000scholars seeks a PhD candidate to advance software security through repository-level vulnerability detection and root-cause analysis using Large Language Models. The project emphasizes RL fine-tuning to improve reasoning on security tasks and aims at automated remediation across complex codebases.

The successful candidate will conduct deep research under supervision, contribute to publications, and collaborate with an international consortium.

Qualifications

• Master’s degree or equivalent required in CS/AI/Cybersecurity or related field.
• Solid knowledge and project experience in designing, training, and evaluating DNN-based classification models.
• Familiarity with static code analysis using CFGs and DFGs.

Responsibilities

• Conduct full-time research centered on the thesis theme.
• Develop RL-based fine-tuning methods to improve security reasoning in LLMs.
• Investigate repository-wide vulnerability discovery and automated remediation concepts.
• Collaborate with international partners and document results for scientific publication.

Skills

Python

Deep Learning

Program Analysis

Tree-sitter

Joern

PyTorch

TensorFlow

NLP

Adversarial ML

Education

Master’s degree in Computer Science, Artificial Intelligence, Cybersecurity, or related

Tools

PyTorch

TensorFlow

Tree-sitter

Joern

Job description

1000scholars seeks a PhD candidate to advance software security through repository-level vulnerability detection and root-cause analysis using Large Language Models. The project emphasizes RL fine-tuning to improve reasoning on security tasks and aims at automated remediation across complex codebases.

The successful candidate will conduct deep research under supervision, contribute to publications, and collaborate with an international consortium.