Founding Research Engineer

As a Research Engineer, you’ll be responsible for measuring and identifying dangerous AI capabilities across various domains: cybersecurity, biosecurity, persuasion and manipulation, model autonomy, and more. In addition to building evaluations and conducting red teaming, you will be responsible for guiding and contributing to our AI safety research and keeping up with the latest developments in the research landscape.

What you’ll be doing:

• Building dangerous capabilities evals from the ground up
• Threat modelling, i.e. understanding feasible pathways to catastrophic harm
• Task design, i.e. creating easily-measurable discrete tasks across a range of difficulties
• Red teaming, i.e. identifying potential weaknesses, vulnerabilities, or unintended behaviors in AI systems
• Finding ways to automate red teaming and model auditing processes
• Keeping up with the latest AI research across a variety of relevant domains
• Utilizing the latest prompt engineering and capabilities elicitation techniques to measure true upper-bound capabilities
• Interfacing directly with external clients, partners, and research collaborators
• Designing scaffolding to let autonomous agents use tools, understand their environment, plan, and use long-term memory
• Communicating findings in detailed qualitative and quantitative reports for different audiences
• Scaling up infrastructure to build and run model evaluations more efficiently

What you bring to the table:

• A deep commitment to AI safety
• Experience building complex software projects across all stages of development: architecture design, prototyping, deployment to production, and long-term maintenance/debugging
• Deep Python experience, including familiarity with popular Python libraries and frameworks such as NumPy, pandas, Flask, asyncio, scikit-learn, TensorFlow, and PyTorch
• Excellent communication and collaboration skills, with the ability to articulate technical concepts clearly to both technical and non-technical audiences
• Grounding in software engineering best practices, including version control, testing, and documentation
• Familiarity with basic web technologies: HTTP, HTML, CSS, and JavaScript
• At least high-level knowledge of common LLM architectures (e.g. GPT, BERT), and experience with fine-tuning and/or prompting LLMs
• Experience with data analysis, visualization, and reporting tools (e.g. pandas, Matplotlib, seaborn, Plotly)
• Some exposure to Docker and related tools (e.g. Docker Compose, Kubernetes)
• Ability to be flexible and wear multiple hats, context-switching between different roles and responsibilities as needed

Why you’ll love it here:

• Make a meaningful impact solving one of the biggest challenges of our lifetimes
• Work from nearly anywhere in the world
• 20 vacation days and unlimited family & caregiver leave
• Regular company off-sites at fun, energising locations
• Inclusive culture that embraces diverse perspectives and backgrounds
• Continuous learning opportunities and clear paths for career growth

Compensation:

• Competitive salary in your preferred locale that reflects your experience and the value you bring to the table. We believe in paying fairly and rewarding top talent.
• Generous stock options that give you a stake in the company's success. Equity ranges from 0.1% to 2% depending on your skills and background.

‍