Want to know if this job is worth applying to?
United States
Remote only
This position is listed on behalf of a partner company, who manages all applications and next steps. Our partner is looking for a Technical Product Manager, Trust & Safety based in the United States.
The Technical Product Manager, Trust & Safety will define and deliver advanced data products that help leading AI research teams train and evaluate safer, more trustworthy models. This role owns the full product lifecycle, translating research needs into product specifications, reinforcement learning environments, safety rubrics, and evaluation frameworks. A central focus is designing realistic, sandboxed scenarios that expose vulnerabilities such as reward hacking, deception, prompt injection, and unsafe tool use. The successful candidate will establish measurable safety and performance standards while ensuring AI agents can complete tasks responsibly under adversarial conditions. Working closely with researchers, engineers, red-teamers, policy experts, and specialist delivery teams, they will turn emerging risks into practical product improvements. This is an opportunity to shape the development of trustworthy agentic AI in a technically demanding, fast-moving environment.
Define and deliver data products for AI research teams by translating research objectives into clear specifications for safety training data, reinforcement learning environments, and evaluations.
Design sandboxed, trap-based reinforcement learning environments that test for misalignment, reward hacking, specification gaming, deception, sandbagging, oversight evasion, and power-seeking behavior.
Develop cybersecurity-focused scenarios involving prompt injection, exposed credentials, credential exfiltration, poisoned tools, and unsafe file or network actions.
Build safety into agentic tasks through behavioral rubrics, verifiers, and reward signals that guide models to complete work while refusing, escalating, or requesting confirmation when appropriate.
Define and manage evaluation frameworks and measurable safety metrics, including attack success rates, over-refusal, trap activation rates, and task completion under adversarial pressure.
Identify evaluation gaps and blind spots, analyze results, and drive continuous product improvements.
Prioritize product opportunities by defining problems, evaluating technical and business tradeoffs, and balancing minimum viable solutions with longer-term goals.
Collaborate with AI researchers, machine learning and platform engineers, red-team specialists, policy experts, and delivery leads managing expert annotator teams.
Monitor developments in agentic AI risks, alignment research, and emerging attack techniques, translating relevant findings into product roadmap priorities.
Prepare clear product specifications, customer briefs, and safety reports for both technical and executive audiences.
Develop reusable trap libraries and quality frameworks that enable delivery teams to create reliable, difficult-to-detect evaluation environments efficiently.
At least 5 years of product management experience, including ownership of technical products in machine learning, AI data, cybersecurity, or trust and safety.
Practical familiarity with large language model training and evaluation methods, including supervised fine-tuning (SFT), reinforcement learning from human feedback (RLHF), reinforcement learning from AI feedback (RLAIF), reward models, verifiers, and benchmark design.
Working knowledge of agentic AI systems, including tool use, computer-use agents, multi-step planning, and associated failure modes.
Strong understanding of AI safety concepts such as reward hacking, deceptive alignment, jailbreaks, and prompt injection.
Technical fluency sufficient to read code, review environment specifications with engineers, and make informed technical tradeoffs.
Familiarity with Python and sandboxed environments such as Docker, virtual machines, and browser automation is an advantage.
Experience designing metrics that measure risk, system performance, and user impact, and using data to guide product decisions.
Demonstrated ability to launch products from 0 to 1 in fast-paced, ambiguous environments.
Excellent written and verbal communication skills, with the ability to explain complex technical concepts to researchers, customers, and executives.
Creative and adversarial problem-solving skills, including the ability to anticipate how advanced AI systems might exploit weaknesses or bypass safeguards.
Annual salary range of $180,000–$220,000 USD, depending on experience, skills, and qualifications.
Remote work opportunity within the United States.
Opportunity to shape the safety and reliability of advanced AI systems.
Cross-functional collaboration with specialists in AI research, engineering, cybersecurity, policy, and evaluation.
