About Primary
Far too few of the world's best journalists get the recognition, audience, or rewards their work deserves. Primary is building the infrastructure to change that. Primary Index is the first AI-native ranking system for journalism — the credibility standard for news in the AI economy. We evaluate individual journalists on the accuracy, depth, and craft of their published work through a purpose-built AI operating against a transparent, published methodology. Building an AI system that can faithfully assess factuality, evidence quality, source reliability, and editorial balance at scale — faithfully enough to earn the trust of journalists and the public — is a genuinely hard problem that requires new methods, not the application of existing ones. That is the work of this team.
We are seeking a Research Engineer to help design, build, evaluate and deploy the next generations of AI systems for automated knowledge acquisition and evaluation.
You will work closely with the Chief Scientist on research and development spanning large language models, automated reasoning, knowledge acquisition, and high-throughput AI systems. Rather than applying existing AI techniques to business problems, you will help invent, implement and evaluate new approaches.
Current projects include:
large-scale local inference over news and other text corpora;
AI systems that assess factuality, neutrality, use of evidence and source quality;
faithful explanation and justification generation;
large-scale evaluation and scoring of published journalism;
adaptation and fine-tuning of open-weight language models;
new methods for evaluating and improving AI systems.
Overview
This is a research-oriented role in which AI-assisted software development is expected. The emphasis is on technical judgement, problem solving and independent research rather than proficiency with particular implementation tools or frameworks.
We value intellectual independence and ambition within a team. We are looking for researchers who enjoy taking loosely specified problems, identifying the key research questions, formulating new approaches, and critically evaluating evidence. The ability to reason well, learn rapidly, understand and make suggestions, and exercise sound technical judgement is more important than familiarity with any particular software framework.
Responsibilities
You will:
translate high-level research ideas into working AI systems;
identify relevant literature and implement or extend state-of-the-art methods;
design and carry out experiments to evaluate alternative approaches;
develop new algorithms, objectives and evaluation methods where appropriate;
effectively employ and critically evaluate AI-based software development tools;
build robust, scalable research prototypes;
communicate results clearly through technical reports and research papers.
Essential qualifications
You should demonstrate:
a strong background across artificial intelligence, machine learning or computational cognitive systems;
the ability to independently design and build sophisticated AI systems;
excellent analytical and experimental skills;
the ability to rapidly understand and extend new research;
strong technical judgement and intellectual independence;
excellent written and spoken communication.
A PhD is desirable but not essential. Outstanding candidates with an MSc or equivalent experience are encouraged to apply.
Evidence of suitability
We are interested in evidence that you can independently transform research ideas into effective AI systems. Evidence of suitability includes one or more of:
first-author or substantial contributor research publications;
substantial AI research projects;
open-source AI software;
novel machine learning methods;
significant technical contributions to research systems;
demonstrated ability to rapidly understand and extend unfamiliar AI systems.
What will distinguish outstanding candidates
Outstanding candidates will be able to take a broad research direction—for example,
“Investigate whether a ranking-based loss, alternative supervision strategy, or architectural modification could improve factuality scoring while producing faithful explanations.”
—and independently:
formulate promising approaches;
identify relevant prior work;
design appropriate experiments;
build and evaluate candidate systems;
interpret results critically;
recommend the most promising direction.
Research environment
You will work closely with experienced AI researchers in an ambitious, high-velocity startup on problems at the intersection of language models, reasoning, knowledge acquisition and trustworthiness. Our objective is to build AI systems that materially advance the state of the art while rapidly delivering real-world impact. Successful work is expected to result in both rapid deployment in production and publication at leading AI conferences.
Compensation & Structure
Compensation: $150,000+ cash compensation plus generous equity package, depending on experience and seniority
New York City–based role, working in person with the research team
Meaningful equity grants and an annual performance bonus
Comprehensive benefits, including health, dental, and vision
We will calibrate seniority and scope to the right candidate