AI Researcher

Reporting to: Head of AI

About Us

Novee is the AI penetration testing platform built for a world where attackers operate at machine speed. Our AI agents think like real attackers, continuously finding and validating exploitable risk so security teams can fix what matters first.

We raised $51.5M in four months, backed by top investors YL Ventures, Canaan Partners, and Oren Zeev - one of the fastest funding trajectories in offensive security. Our proprietary AI models outperforms frontier LLMs on real exploitation benchmarks, and we're just getting started.

We're a global team with offices in Tel Aviv, the US and the UK, founded by offensive security experts. We work with hundreds of global enterprises to help them secure their environments and innovate faster.

We're at the cutting edge of AI and cyber. We move fast, work with the best, and have fun together.

About The Role

We’re looking for an AI Researcher to join our model training team, which designs, trains and evaluates the language models powering Novee’s autonomous pentesting agents.

This role is a unique opportunity to help push the frontier of AI capabilities in cybersecurity. Reporting to our Head of AI and working alongside exceptional researchers, you will develop and apply cutting-edge reinforcement learning methods to shape the behavior of our models in dynamic, long-horizon environments.

You’ll own the full research lifecycle: Defining hypotheses, designing experiments, training models, analyzing results and iterating. You’ll work across the full post-training stack, including data pipelines, verifiers, reinforcement learning algorithms and evaluations.

This role combines deep research with hands-on engineering, and has a direct impact on the quality of our product.

Responsibilities:

• Research, train, and evaluate language models for cybersecurity tasks.
• Devise evaluations that capture what matters. Make sure the numbers are meaningful and make the numbers go up. Analyze errors and failure modes.
• Develop post-training recipes to improve model reasoning, tool use, exploration and decision-making in challenging offensive security tasks.
• Develop large scale data pipelines, complex RL environments, reliable verification methods and effective reinforcement learning algorithms.
• Design, run and analyze experiments to prove new methods and ship compounding improvements.
• Collaborate with cybersecurity researchers on training data, evaluations and reward signals.
• Partner with our infrastructure team to rapidly scale research efforts.
• Implement research ideas in Python, leveraging modern deep learning tools such as inference frameworks (vLLM, SGLang, …), training frameworks (verl, slime, …), and training APIs (Tinker and similar) in an AI-first development environment.
• Work with engineering teams to translate research outcomes into capabilities used by our agents.
• Review relevant research and assess its practical value for the problems we’re solving.
  
  

• Master’s degree with a research thesis in Computer Science / Mathematics / Physics / Engineering / another STEM field; or Bachelor’s degree plus equivalent industry / army research experience.
• Hands-on experience in data-driven computational research: machine learning, algorithm development, numerical analysis, data science, statistical modeling or similar.
• Proven ability to take open-ended problems from vague ideas to working implementations.
• First-principles thinking and the ability to critically evaluate experimental data and distinguish real effects from noise and bugs.
• A strong understanding of deep learning and language models.
• Comfort with ambiguity and fast-paced research environments where priorities and core assumptions evolve quickly.
• Great communication skills and the ability to collaborate in a small, agile team.
• Independence, curiosity, and a deep bias for action.
  
  

• PhD in a relevant field or relevant research publications.
• Specific background in cybersecurity.
• Specific background in reinforcement learning, particularly for language models or agents.
• Hands-on experience with deep learning training and/or inference frameworks.
• Hands-on experience running and debugging large-scale distributed systems.