Research Unit Delivery Manager - AI Safety Institute

AISI is an incredible and exciting professional challenge. Our team is a diverse and multidisciplinary group from a broad range of personal and professional backgrounds. The successful candidate will be working at the forefront of AI Safety, making significant personal impact on some of the most difficult problems in the field, and ensuring that this transformative technology is harnessed for good. If you want a job with high agency, high impact, and exceptional opportunities to develop and grow then AISI is the perfect place for you.AISIs Research Unit is a dynamic blend of technical researchers and delivery staff who work collaboratively to scope and deliver our research projects. Our delivery team members are integral to our success. They are the driving force behind our research, unblocking issues, managing risk, and ensuring our research remains high quality and in line with the standards demanded by AISI and government. They set the conditions for success for our technical teams. We arent looking for a project manager, you dont need to be fully agile accredited, but rather a pace setter, fixer, and doer. If you can demonstrate that you have delivered big things in complex environments, work collaboratively in multidisciplinary teams, and can operate under pressure, then you should apply. We are recruiting for a Delivery Manager in each of the following teams (six posts total). If you have a preference for specific roles, please make this known in your personal statement. Safeguards The AI Safety Institutes Safeguard Analysis team aims to understand security threats to AI systems, and interventions to mitigate them. As AI systems become more advanced, and integrate in society, the risk of abuse by bad actors will increase. Companies developing frontier AI systems use a variety of techniques (safeguards) to prevent harm caused by malicious users and unreliable or misaligned models. For example, deployed AI chatbots typically refuse explicitly harmful requests, like instructions for building bombs, due to the safeguards put in place.   The Safeguard Analysis team researches such interventions, evaluating protections used to secure current frontier AI systems, and considering what measures could and should be used to secure systems in the future. The scope of bad outcomes we are considering is broad, ranging from threats like prompt injection attacks (jailbreaks) and data poisoning, to threats from traditional cybersecurity attacks. The variety of safeguards for these bad outcomes we are looking to analyse is similarly diverse, including safety training, input/output filtering, retrospective analysis on saved user interactions, and traditional software defences.   The Delivery Manager in this role will be responsible for helping to explore an ambitious, expanded scope for the workstream and then managing the day-to-day delivery of the work. This will include building partnerships across government, the private sector, and academia. Candidate applying for a role within the Safeguards team should be willing to go through DV clearance. Academic Engagement This team seeks to build a bridge between AISI and academia, to push forward foundational safe AI research beyond evals, and beyond research that we can do within government, with a focus on developing new AI mitigations and the next generation of evals. The aim is to get the UK academic community to be aware of opportunities and gaps in research, surfaced by our work, and then drive this forward with support from AISI including compute, funding, and AISI-provided expert advice. The programme seeks to achieve translational impact by communicating AISI views outwards, which galvanises the academic community to work on these areas. We anticipate the work to have at least three streams: i) inviting academics into AISI to give talks and work with the team, ii) active collaborations, to push forward research on AI Safety, and iii) workshops with joint participation from AISI, academia, industry across international boundaries which aim to discuss or identify novel research directions and foster collaboration in the AI safety community. The Delivery Manager in this role will be responsible for scoping and delivering the programme, including: initiating and managing research collaborations, establishing appropriate funding and other support streams, and organising and delivering workshops. You will build strong relationships with the academic community and work across all of AISIs research teams to identify and deliver collaboration opportunities of benefit to them. Systemic Safety (x2) Systemic AI safety is a field that aims to understand and mitigate the broader societal risks associated with AI deployment, beyond the capabilities of individual models. Systemic AI safety focuses on risks and mitigations in the context of AI deployment, both in specific sectors and across society. For example, we want to know what risks could emerge when frontier AI is integrated into education, healthcare and finance such research requires understanding of both technical aspects of AI, but also human and social aspects of the sector-specific context. To date, AISIs Systemic Safety team has been focussed on delivering Phase 1 of our Systemic Fast Grants programme. This team will be focussed on identifying and catalysing interventions which could advance the field of AI safety and strengthen the systems and infrastructure into which AI systems are deployed. This team will lead the effort to shape the AI ecosystem to ensure safe, scalable, and responsible AI development, and will work with the private sector, academia, and model developers to develop actionable technical innovations that governments or industry could implement. The Delivery Manager(s) in this team will be responsible for helping to explore an expanded scope for the workstream and then managing the day-to-day delivery of the work. This is likely to include, at a minimum, identifying and delivering novel external partnerships opportunities for example future grants programmes, but a core part of the role will be working to define a maximally ambitious scope for the workstream working alongside our Research Director and other senior colleagues and then translating that vision into tangible delivery. Engineering & Digital (x2) Delivery Manager(s) covering these areas play an enabling role that impacts all of AISIs technical work. It is not a requirement for candidates to have software engineering skills or qualifications; however, experience delivering ambitious, multi-disciplinary projects in a technical context is strongly desirable. These roles will facilitate the work of AISIs engineers and wider researchers, both proactively and reactively. As a technical organisation delivering research on advanced AI models, AISI has developed an outstanding software engineering function. AISI engineers have built a bespoke research platform that enables scalable, automated evaluation of models through API-based model access. The engineering team has also made significant contributions to the wider AI safety ecosystem through projects such as Inspect, an open-source AI evaluations framework available to all via GitHub. Delivery Manager(s) working directly with AISI engineers will establish high impact work with external collaborators, identify and deploy software to meet research requirements, establish robust KIM and team coordination mechanisms and spearhead new projects playing a coordinating role across teams. All of AISIs technical work is enabled by having the right digital services including hardware, software and compute. As AISIs research activity scales and evolves, Delivery Manager(s) in this area will play an essential role in working with technical staff to identify requirements, designing AISI's internal governance mechanisms to manage and deploy technology, and integrating with broader corporate services across a range of fast-paced and novel digital projects. Responsibilities of a Delivery Manager Amongst other things, your responsibilities will include: Playing a cross-cutting strategy and delivery role within your workstream. You will work with the Workstream leads, other researchers, and other delivery staff to deliver multiple research project and help shape the long-term strategy of the team. You will have a key role in setting out the strategy of and delivering the workstream. Generating and maintaining delivery momentum within AISIs Research Unit by creating a culture of velocity, coordination, and tight execution cycles. You will oversee delivery of high-quality research projects conceptualised by our technical team and external partners. You will identify and clear roadblocks, set a collaborative and productive team culture, and continually strive for increased efficacy with high-quality delivery as your number one priority. Acting as a key point-person for your workstream. As the Grade 7 Delivery Manager in this team, you will be across the full breadth of the portfolio and able to speak confidently to its aspects. Create and manage an ambitious portfolio of research partnerships, which may be delivered through grants, competitions, and other arrangements, and then leverage these partnerships to achieve real-world impact. Play an active role in creating a culture of innovation and exploration, encouraging colleagues to take managed risks, fail fast, and learn. You will have a key role in making sure AISI is managing risk appropriately whilst minimising any unnecessary bureaucracy.  Delivering through others: much of your time will be spent on cross-team projects. You will need to collaborate across boundaries, bringing together virtual teams to deliver.   Line management of at least one member of staff. With the potential to scale this over time. Building and leveraging a network of partners and stakeholders across Government and externally, and insight and outside ...