Capable and Reliable

Two requests can look nearly identical and be worlds apart. "See if you can find me an example of this" needs a capable system: it finds the example or it doesn't. "Conduct a reasonable search for any and all documents responsive to this request" is a different kind of promise. Its answer spans a corpus no one will ever read end-to-end. So the system's process, as much as its output, has to earn the trust of the professionals who rely on it.

That property is reliability. It decomposes into consistency, robustness, calibration, and safety: systems that behave tomorrow the way they did today, degrade predictably under stress, know how confident they should be, and check their own work. Before aiR returns an analysis, it validates its citations and runs internal consistency checks; when a check fails, it refuses to answer. It has refused more than a million times so far in 2026, and we count every one as a success: an error caught before it reached a user.

You'll build for both, and set the direction for how your workstream does it.

The Focus: Public Sector Investigations

This role anchors our science for public-sector investigations: FOIA and public-records workflows, government investigations, and the agencies that run them. Records officers and investigators face the same combinatorial problem as litigators, with an added obligation: their results answer to the public. You'll own the science for this domain end-to-end, from retrieval and review systems to the evaluation standards that make responses complete, correct, and defensible.

What You'll Do

• Set technical direction for the workstream: choose the approaches, sequence the science, and keep the work coherent across contributors.
• Lead the science without managing the people: review other scientists' designs and evaluations, raise the bar, and unblock the hard problems.
• Own AI system readiness across the workstream, from problem framing through evaluation, error analysis, efficacy studies, and production monitoring, in partnership with engineering.
• Be the first escalation for the domain's hardest technical calls, and know when to bring in Staff and Principal partners.
• Be a pathfinder: build the first prototypes with tiger teams, and carry early initiatives from idea to evidence.
• Take the work from proof-of-concept to production at scale, working with product, engineering, design, and the legal experts embedded in the team.

What You Bring

• A master's or PhD in computer science or another quantitative discipline (or equivalent professional experience), and at least 3 years of professional experience in applied AI/ML.
• Eligibility for a US Public Trust clearance.
• Working knowledge of privacy and records standards: FOIA exemptions, PII handling, and redaction.
• Deep applied AI/ML and deployment experience: AI systems you specified, prototyped, and carried into production with engineering partners.
• Enough depth in modern AI to direct others' work with it: you can tell a sound AI system from a fragile one before it ships.
• You lead the data understanding for your workstream: you design the study that settles the argument, and you trust data over intuition.
• Software-engineering judgment that sets a workstream's tone: your reviews teach, and your designs last.
• An ownership mindset that extends beyond your immediate scope.

Nice to Have

• Deeper experience with government data or public-records work.
• An interest in legal technology and the justice system.
• Experience developing information retrieval systems.
• Experience developing agentic harnesses.
• Experience building reliable AI systems at scale.

Why Relativity Applied Science

This is the place where your curiosity, dedication, and talent will build products that power the pursuit of justice around the world.