Research Engineer – AI Advantage Evaluation

Context: The CCTI

The Centre for Cryptographic Trust Infrastructure (CCTI) is part of the ARIA-funded opportunity space "Trust Everything, Everywhere," which aims to enable AI agents to successfully coordinate with others despite competing objectives, information asymmetry, and adversarial environments.

CCTI’s role in this opportunity space is to enable AI agents to find counterparties, negotiate terms, and execute real-world transactions without the prior trust history that humans depend on. CCTI does this by making agents’ claims verifiable, both about the real-world entities they represent and about their reasoning.

CCTI’s north star is to build the infrastructure for such proofs, and to investigate whether, and demonstrate that, this infrastructure unlocks coordination among AI agents.

CCTI is a consortium of the University of Cambridge (Co-PI Dr. Martin Kleppmann, Associate Professor), SINE Foundation (Co-PI Martin Pompéry), Ink & Switch, and Light Squares.

Research context

CCTI operates in the broader research area of Cooperative AI (https://arxiv.org/abs/2012.08630), which studies settings in which agents must coordinate despite mixed motives and potentially non-overlapping incentives and interests.

CCTI’s infrastructure relies on cryptography that keeps data (about the real world) confidential while still allowing it to be processed, for example to produce cryptographic proofs, without necessarily disclosing the underlying data. We are building cryptography-enabled coordination primitives, such as matchmaking among agents, that use trusted execution environments (TEEs) or secure multi-party computation (MPC) to enable coordination over confidential (input) data.

In addition, we apply game-theoretic research to the design of our infrastructure, coordination primitives, and frameworks, so that AI agents can cooperate where they otherwise would not.

One of our key objectives at CCTI is to research the “AI advantage” of our infrastructure, defined as its effect on cooperation compared with (a) a human-only baseline and (b) AI agents without access to CCTI infrastructure and primitives.

The position

Title: Research Engineer – AI Advantage Evaluation

Term: full-time (remote-only), start as soon as possible; fixed-term until December 2027; one-year extension contingent on a follow-on funding decision; remote possible (EU time zones)

Your role

You lead the measurement and evaluation work at CCTI. You design the methodology to evaluate whether AI agents with CCTI trust infrastructure reach agreements where agents without it—and humans—fail. Together with the CCTI team, you co-design and implement the operationalization of CCTI’s main frameworks and infrastructure for and by AI agents.

What you will build

• You will develop the methodology for measuring the AI advantage.
• You will build an evaluation harness that operationalizes this methodology and reports null results as readily as positive ones.
• You will create bridges between CCTI’s cryptographic trust infrastructure, its game-theoretic frameworks, and (LLM-based) agents.
• You will deliver prototypes and experiments grounded in real-world use cases that demonstrate agent behavior, agents’ interactions with CCTI infrastructure, and measured differences in outcomes.

Profile and background

Essential (in order of importance):

1. LLM-agent engineering, including agent skill design, eval loops, and tool use. You have built agents with scientific rigor: agents that run unattended and reproducibly, with pinned model versions and archived prompts and transcripts, so that any run can be repeated and audited.
2. Experimental design and execution. You have designed experiments yourself, including controlling for confounders and conducting power analyses. You can pre-register a design, implement it faithfully, run it at scale, and carry out the pre-specified statistical analysis correctly (effect sizes, confidence intervals). You notice when an implementation choice threatens the validity of the design, and you address it instead of working around it.

Nice to have:

• You have led a measurement effort that meaningfully changed the trajectory of a research project.
• You are familiar with information economics or transaction-cost economics.
• You have a demonstrable background in interdisciplinary research.

Not required:

• a background in cryptography, or experience running studies with human participants.

What we offer

1. The chance to develop and publish a measurement methodology for an emerging field: AI advantage in trust-mediated cooperation.
2. The opportunity to collaborate with leading innovators and researchers at the University of Cambridge and across the ARIA opportunity space.
3. Everything you build is open source by default (MIT/Apache-2.0), and all papers are released as preprints. The work is funded by ARIA, with publications expected in 2027.
4. Salary: €65,000–75,000 per year; either regular employment or a freelance contract is possible, depending on your tax residence.

How to apply

Please send your CV, together with either a cover letter or links to one or two pieces of work that best demonstrate the essential requirements above (e.g., a repository, a paper, etc.), to jobs@sine.foundation.

Application deadline: October 7, 2026

About SINE

SINE is a Berlin-based nonprofit organization that builds digital public goods for collective action problems, such as open-source interoperability standards, governance frameworks, or cryptographic protocols for multi-party computation. Our focus has always been on why organizations fail to coordinate even when doing so would leave all of them better off, and on how to unlock inter-organizational cooperation for the common good.

SINE is an equal opportunity employer. We are committed to equal employment opportunity regardless of race, ethnic origin, sex, gender identity, religion or belief, sexual orientation, age, marital status, or disability. Please do not submit personal data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, trade union membership, data concerning your health, or data concerning your sexual orientation.

SINE processes the data provided in your job application in accordance with our Recruitment Privacy Policy ( https://legal.sine.foundation/recruitment-privacy-policy/ ).