Technical Product Manager, Trust & Safety

Talenthop; United States; On-site; USD 180,000 - 220,000; Full time

Job summary

Talenthop is seeking a Technical Product Manager, Trust & Safety to define data products enabling safe and reliable AI model training and evaluation. You will translate research needs into specifications and design RL environments and assessments to mitigate risk.

You will collaborate with researchers, engineers, policy experts, and delivery teams across our global labs and customers, shaping the roadmap and communicating outcomes to executives.

Qualifications

• 5+ years of product management with ML or security products.
• Experience with large language model training and evaluation.
• Ability to explain complex AI concepts to diverse audiences.
• Experience designing risk, performance, and impact metrics.
• Proven track record delivering new products in uncertain environments.

Responsibilities

• Define data products based on lab research goals for safety training and evaluations.
• Design sandboxed reinforcement learning environments to test AI behavior.
• Specify safe-behavior rubrics, verifiers, and reward signals.
• Define and track evaluation metrics for safety and trust.
• Prioritize development tasks by business and technical tradeoffs.
• Collaborate with researchers, engineers, policy experts, and delivery teams.
• Monitor threat landscape and incorporate risks into roadmap.
• Communicate specifications and findings to technical and executive stakeholders.

Skills

ML PM

LLM training

AI safety concepts

Python familiarity

Metrics design

Adversarial thinking

Stakeholder communication

Tools

Sandboxed environments

Job description

The organization operates in the global data engineering industry, focusing on the advancement of artificial intelligence (AI). It addresses the challenge of building trusted AI systems at scale by providing data, evaluation frameworks, and human expertise necessary for AI development. With over 36 years of experience, the company delivers high-quality data solutions and services tailored for Generative AI and AI adopters.

About the Opportunity:

The TECHNICAL PRODUCT MANAGER, TRUST & SAFETY is responsible for defining and delivering data products that enable frontier labs to train and evaluate AI models for personalization and other agentic tasks safely and reliably. The role involves managing the full product lifecycle from translating research needs into specifications to designing reinforcement learning environments and evaluations, ensuring the safety and trustworthiness of AI behaviors. This position contributes by developing technical solutions that mitigate risks in AI model behavior and improve safety outcomes.

Responsibilities:

• Define data products based on lab research goals, including safety training data, reinforcement learning environments, and evaluations.
• Design sandboxed reinforcement learning environments embedding misalignment and cyber traps to test AI model behavior.
• Specify safe-behavior rubrics, verifiers, and reward signals to guide models toward safe task completion.
• Define and track evaluation metrics for safety and trustworthiness, identifying blind spots and iterating accordingly.
• Prioritize development tasks by evaluating business and technical tradeoffs.
• Collaborate with researchers, engineers, red-teamers, policy experts, and delivery teams.
• Monitor the threat landscape for emerging risks and incorporate them into the product roadmap.
• Communicate specifications, findings, and briefs clearly to technical and executive stakeholders.

Requirements:

• Minimum of 5 years’ product management experience with technical products in machine learning, AI data, security, or trust and safety.
• Practical knowledge of large language model training and evaluation methods including reinforcement learning techniques.
• Understanding of agentic AI capabilities and associated failure modes.
• Strong grasp of AI safety concepts such as reward hacking, deceptive alignment, and prompt injection.
• Technical proficiency sufficient to review code and environment specifications; familiarity with Python and sandboxed environments is advantageous.
• Experience designing metrics to measure risk, performance, and impact, using data to guide decisions.
• Proven track record of delivering new products in fast-paced, uncertain environments.
• Ability to clearly explain complex technical concepts to diverse audiences.
• Creative and adversarial thinking to anticipate potential model exploits beyond existing checklists.

Pay Range and Compensation Package:

• The expected salary range for this position is $180,000 – $220,000 USD per year, based on experience, skills, and qualifications.

Equal Opportunity Statement:

Our client is an equal opportunity employer. They celebrate diversity and are committed to creating an inclusive environment for all employees. All qualified applicants will receive consideration for employment without regard to race, color, religion, gender, gender identity or expression, sexual orientation, or national origin.