Technical AI safety: Infrastructure Engineer – now closed

Deadline: Friday 7th of August, applications are reviewed on a rolling basis, but it is advantageous to apply as soon as possible.

Job title: Infrastructure Engineer (Contractor)

Point of Contact: Beth Nichols

Supervisor: Tommy Shaffer Shane

Salary: £700–£800 per day, depending on experience.

We’re hiring this as a contract role to get going quickly, there may be opportunity for this role to turn into a permanent position at the end of the contract, pending further assessment. 

Hours: Flexible, initially a 6 month contract, averaging 2 days per week. 

Location: This is a remote, flexible position with no set work hours or workplace. However, successful candidates will be entitled to use our fantastic office space in Whitehall, London and encouraged to meet the team to gain a thorough understanding of our work.

Context

The Loss of Control Observatory is a world-first capability for detecting and monitoring uncontrolled, misaligned AI systems. Our ambition is to become the world’s largest and most authoritative evidence base for loss of control, through real-time monitoring of rogue agent behaviours and threat indicators to inform government and AI lab action.

We have so far launched a successful UK AISI-backed pilot, using social media analysis, which detected a 5x increase in loss of control incidents in 2026 and was covered in international news. We now plan to scale this pilot into a robustly engineered and scalable platform, integrating new data sources and enabling world-leading research on one of the most important global problems.

Your role

We require an infrastructure engineer to realise our ambitious vision for the Observatory. You would overhaul the existing data pipeline, classification system, and monitoring dashboard. And you would scale the system with additional and larger data sources and improved data collection methods.

This is a technical development role focused on maintaining the Observatory’s core infrastructure (data collection pipelines, LLM-based classification systems, and a web-based monitoring dashboard) and extending it to cover additional data sources, longer collection windows, and new risk areas. You’ll work with CLTR’s AI Unit to scope expansions and ensure the platform reliably supports world-leading monitoring work.

What you’ll do:

• Refine and maintain our Python data pipelines for collecting and processing social media data (X, Reddit, and potentially others)
• Build pipelines for scraping and analysing chatbot share links
• Implement GDPR compliant media artifact storage so evidence is retained even if source platforms delete originals
• Maintain and refine LLM-based classification systems, including prompt engineering and evaluation frameworks
• Build and maintain the access layer for external parties – warehouse schema, the monitoring/review dashboard, self-service analytics, exportable datasets and, over time, an API.
• Deploy and maintain cloud-based services with automated/scheduled operations
• Implement robust error handling, logging, and data validation throughout the system

Requirements

Essential

• Strong Python skills for data processing and ML/LLM workflows
• Experience interfacing with APIs (social media platforms, LLMs, cloud storage) and data scraping
• Strong interpersonal and communication skills, with experience of working effectively with non-technical colleagues
• Web development skills (dashboard building, authentication, WSGI deployment)
• Strong practices around testing, error handling, and version control (Git)n control (Git)

Desirable

• Experience with prompt engineering and LLM evaluation
• Familiarity with and interest in AI safety
• Understanding of AI risks and behaviours relevant to the project (e.g. loss of control, scheming)
• Experience with OSINT methodologies or social media data analysis

We are particularly interested in hearing from people who…

• Have built infrastructure or software in industry or big tech and want to bring those skills to AI safety – bonus if you’ve done that already. 
• If you have spent time building and scaling real systems, and you care about the loss-of-control problem but don’t want a research role, this could be a great fit. 
• You don’t need a research background – we’re looking for strong builders who want to help take the Observatory to the next level. 

How to Apply

Email beth@longtermresilience.org by 9am, Friday 7th of August with the following items:

1. Your CV

2. A link to your most relevant GitHub repository — ideally something that gives us a sense of how you build and scale real systems. If your strongest work is private or confidential, then please include a short description of what you built, your role in it, and the technical choices you made instead.

3. A short note (~250–300 words) on what draws you to this work — we’re interested in your motivation and how you communicate.

As an employer, we encourage candidates from all backgrounds to apply and do not discriminate based on age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex, and sexual orientation.

If you have any questions, please reach out to beth@longtermresilience.org

Use of AI in our recruitment process
We use AI-assisted tools to support the initial sifting of applications. These tools help us assess CVs and application materials against the criteria set out in this job description. AI outputs are reviewed by a member of the CLTR team before any shortlisting decisions are made — no purely automated decision-making takes place. For more information on how we process your personal data during recruitment, including your rights, please see our candidate privacy notice. If you have questions about how AI has been used in assessing your application, you can contact beth@longtermresilience.org.