As AI adoption increases, the challenge in the next decade is to harness AI safety with the appropriate levels of governance and assurance to build trust in these advanced algorithmic systems. Most enterprises do not have the capability to do this and we are a new AI assurance venture that provides services to validate AI systems for accuracy, robustness, explainability, fairness, privacy and security.
We are looking for a senior data scientist with experience in deep learning based language models to work with a category-defining AI assurance venture that will help companies test and audit their AI systems. You will help evaluate and stress-test AI models to make sure they are fit for purpose and safe to be deployed. We value strong technical ability and real world experience and there will be room to solve challenging problems and adopt cutting edge technology into business applications.
YOU WILL
- Perform technical AI evaluations, benchmarking and red-team tests on large language models, including assessing them for robustness in performance, embedded biases, vulnerability to jailbreaks and prompt injection attacks.
- Work with the product management team to develop a suite of technical and analytical AI evaluation frameworks and tools that are backed by scientific research and methods. These should assess the robustness, explainability, fairness, privacy, safety and security of AI and. machine learning systems, with a strong focus on large language models.
YOU ARE ABLE TO
- Think from first principles and want to tackle the most challenging technical problems from a multi-disciplinary approach e.g., design, engineering and social science.
- Lead by example regardless of whether you are a manager or individual contributor. You want to work with passionate and talented individuals and people want to work with you.
- Communicate in an open, frank and respectful manner.
- Thrive in a fast-paced environment.
- Navigate uncertainty by being willing to explore, while remaining laser focused on the mission at hand.
YOU HAVE
- Extensive experience as a data scientist training and deploying deep learning based natural language models/large language models in real-world contexts. At least 6 years of workingexperience or a relevant postgraduate degree with at least 4 years of working experience.
- Familiarity with evaluation approaches for language models, preferably with a deep understanding of the failure modes of LLMs including robustness, jailbreaks, embedded biases and possible alignment problems.
- Ability to be independent productive working with Python-based notebooks and to execute on basic data engineering and visualisation tasks (eg. SQL, using StreamLit to build visualisations).
- Passion and interest in applied research on the safe and responsible use of AI and with large language models.
WHO WE ARE
- We are part of a mission driven venture that wants to make AI safer for the world.
- We believe that algorithmic decision making has a huge potential to drive change. However, the increased complexity poses risks that need to be deliberately managed. We want to partner with a broad range of enterprises that have ambitious uses for artificial intelligence and are also serious about utilising AI in a trustworthy manner.
- We are a team of first-principles, multi-disciplinary thinker-doers who believe in doing things right. We are thoughtful about technology and its potential pitfalls. We want to systematically support businesses to build safe, effective and impactful AI technology.
