Senior Machine Learning Engineer - LLM Guardrails

Company Description

Full Court Defense builds real-time security infrastructure for AI agents and LLM-powered applications. Our platform protects production AI systems from prompt injection, jailbreaks, data exfiltration, credential theft, policy bypasses, and other emerging LLM-specific attacks.

At the core of our product is a multi-layer detection engine combining fast rule-based defenses, ML classifiers, specialist models, semantic similarity, and LLM-based judging to analyze both user inputs and model outputs in real time. The platform integrates with any LLM provider in minutes, helping companies deploy AI applications safely without slowing down their product teams.

Headquartered in Tel Aviv, Israel, Full Court Defense is building the security layer for the next generation of AI-native software.

We are looking for a full-time Senior Machine Learning Engineer specializing in LLM Guardrails to join our hybrid team in the Tel Aviv District, Israel, with flexibility to work from home.

In this role, you will lead the development, optimization, and deployment of ML systems that detect and prevent attacks against AI agents and LLM applications. You will work on real-time classifiers, adversarial robustness, multilingual detection, security-focused evaluation datasets, model calibration, and production inference pipelines.

You will play a central role in improving our detection engine, researching new threat vectors, designing scalable ML algorithms, and working closely with engineering and product teams to ship security features used in production environments.

• Develop and improve ML models for detecting prompt injection, jailbreaks, data theft, social engineering, policy bypasses, and other LLM attack patterns.
• Build scalable real-time inference pipelines optimized for latency, accuracy, and cost.
• Create and maintain high-quality evaluation datasets, regression suites, and adversarial test cases.
• Research new LLM vulnerabilities and translate them into practical detection mechanisms.
• Improve model robustness against obfuscation, multilingual attacks, encoding tricks, adversarial suffixes, and prompt-level evasion.
• Collaborate with backend and product teams to deploy models into production security workflows.
• Analyze false positives and false negatives from real-world usage and use them to improve system performance.
• Contribute to the architecture of a multi-layer guardrail system combining ML, statistical methods, rules, embeddings, and LLM-based reasoning.

• Strong experience with machine learning, pattern recognition, neural networks, and statistical modeling.
• Solid computer science fundamentals, including algorithms, data structures, and scalable system design.
• Strong Python programming skills and experience with ML frameworks such as PyTorch, Transformers, scikit-learn, or equivalent.
• Experience building production ML systems, including training, evaluation, deployment, monitoring, and iteration.
• Familiarity with LLM architectures, prompt injection, jailbreaks, adversarial ML, or AI safety is a strong advantage.
• Experience with NLP, text classification, embeddings, model calibration, or multilingual models is a plus.
• Ability to design efficient algorithms for large-scale text processing and real-time decision systems.
• Degree in Computer Science, Data Science, Machine Learning, or a related field, or equivalent industry experience.
• Strong communication skills and ability to work in a fast-moving hybrid team environment.