Jobs at Apollo Research

Execute Apollo Research’s day-to-day finance operations, including procure-to-pay and order-to-cash cycles, invoice and payment processing, overdue balance follow-up, transaction recording in Xero, and weekly multi-entity, multi-currency bank reconciliations. The role requires strong bookkeeping knowledge, finance-platform proficiency, meticulous accuracy, integrity, initiative, and at least two years of transactional processing experience. It offers opportunities to develop toward controllership or FP&A.

Founding DC office staff will communicate AI safety research, build relationships with policymakers, develop policy strategies, monitor legislation, draft policy submissions and memos, and brief senior leaders. Roles span science communications, stakeholder engagement, and policy/legal research. Candidates should understand U.S. government and intelligence-community decision-making, support high-stakes engagements, and help advance safe and secure AI governance. The position is full-time in Washington, DC, with at least 25% travel, including travel to London.

Design, implement, and operate Apollo’s security controls across identity, cloud infrastructure, endpoints, and incident response. Build secure-by-default systems, evaluate security vendors, lead security projects, shape the roadmap, and promote effective security practices across the organization. Key initiatives include zero-trust migration, agentic SOC development, sandbox hardening, third-party access reviews, incident response, and penetration-testing tooling.

Secure Apollo’s internal software, infrastructure, and AI-agent environments through threat modeling, red teaming, attack-trajectory development, adversary tracking, and detection engineering. Own security findings through remediation by designing durable controls, automated tests, CI/CD integrations, and monitoring. Collaborate closely with engineers and research teams to address novel AI-agent and insider-risk threats while documenting threat models and failure modes clearly.

Lead and grow a product engineering team building Watcher, a coding-agent security product. Responsibilities include planning delivery, managing engineers, hiring, coordinating research and engineering handoffs, handling customer integrations, making architecture decisions, reviewing code, and shipping production code primarily in Python and TypeScript. This player-coach role also establishes engineering processes, protects team focus, and uses AI-assisted tools extensively.

Own customer deployments, integrations, technical evaluations, proofs of concept, and white-glove support for Watcher, a coding-agent security product. Build integrations with coding agents, SIEMs, identity systems, and CI/CD pipelines across cloud, on-premises, and self-hosted environments. Troubleshoot customer issues, tune security monitors, communicate with technical and security leaders, and convert recurring customer needs into reusable product capabilities, documentation, and deployment tooling.

Build and operate Watcher, a full-stack coding-agent security product. Develop backend APIs, ingestion and grading pipelines, agent hooks, CLI tooling, monitoring dashboards, enterprise integrations, and real-time blocking systems. Own features from customer discovery through production deployment, while ensuring reliability, scalability, tenant isolation, encryption, access controls, and flexible cloud, on-premises, and local deployments. Collaborate with customers and security teams to improve the product and productionize monitoring research.

Conduct empirical research into AI "scheming": design and run RL experiments on LLMs, build model organisms, develop scalable evaluation techniques, study scaling laws and AI cognition, and collaborate with partner labs to influence deployment and mitigation strategies.

Design, run, and automate pre-deployment evaluations of frontier AI models. Analyze large model transcripts to surface behaviors, build scalable evals, collaborate with AI labs, and improve evaluation pipelines and tooling.