Senior Responsible AI Researcher

Alice (Formerly ActiveFence); San Francisco (CA); On-site; USD 150,000 - 230,000; Full time

Job summary

Alice (Formerly ActiveFence) seeks a Senior Responsible AI Researcher to lead frontier AI safety research at the intersection of model evaluation, Trust & Safety, and adversarial robustness. You will design evaluation frameworks and benchmarks, including Agentic Safety Gyms, to measure capabilities and risks across generative AI modalities.

You will drive self‑initiated research, collaborate with top academic and industry partners, and represent the organization at major conferences, shaping the

Qualifications

• PhD in a relevant field with strong peer‑reviewed publications.
• Proven research record with top‑tier venues.
• Deep domain expertise in AI safety or Responsible AI.

Responsibilities

• Design, build, and deploy new evaluation frameworks, benchmarks, and Agentic Safety Gyms to measure model capabilities and safety.
• Develop adversarial prompting methods, red‑teaming, and behavioral suites to stress‑test LLMs, multimodal systems, and agents.
• Study risks including persuasion, manipulation, bias, and jailbreaking across multimodal and autonomous architectures.
• Lead collaborations with top academic and industry partners (e.g., OpenAI, Google, Anthropic, Microsoft, Nvidia).
• Represent the organization at conferences and safety summits, publishing groundbreaking findings.

Skills

Self-starting research

Academic publishing

Communication

Education

PhD in Computer Science/ML/AI Ethics or closely related

Tools

Python

PyTorch

Transformers

Job description

We are seeking a visionary, highly accomplished Senior Responsible AI Researcher to lead state-of-the‑art research initiatives at the intersection of frontier models, Trust & Safety (T&S), model evaluation, and adversarial robustness.

As part of this role, you will be responsible for leading new frontier evaluation and benchmarking projects every six weeks. Successful candidates are expected to drive self‑initiated research, design evaluation methodologies for emerging risks (e.g., persuasion, manipulation, sycophancy, agentic safety), and build safety benchmarks across Generative AI modalities (Text, Image, Multimodal, and Autonomous Agents).

Key Responsibilities

1. Frontier AI Safety & Evaluation Research

• Design, build, and deploy new evaluation frameworks, benchmarks, and "Agentic Safety Gyms" to measure model capabilities, vulnerabilities, and alignment across diverse risk domains (e.g., Hate Speech, Misinformation, Elections Preparedness, IP/Copyright, Child Safety).
• Develop advanced adversarial prompting methodologies, red‑teaming strategies, and behavioral suites to stress‑test Large Language Models (LLMs), Multimodal systems, and AI Agents.
• Investigate emerging safety risks, including model persuasion, manipulation, sycophancy, bias, and jailbreaking tactics across multimodal and autonomous architectures.
• Lead high‑impact collaborative research projects with top‑tier academic institutions and industry partners (e.g., OpenAI, Google DeepMind, Anthropic, Microsoft, Nvidia).
• Represent the organization as a thought leader and speaker at top‑tier international conferences, technical AI safety summits, and industry events.
• Publish groundbreaking research findings in top AI and safety venues.

3. Cross‑Functional Collaboration

• Partner closely with Product, Engineering, Go‑To‑Market (GTM), Policy and Delivery teams to translate research discoveries into scalable safety products, client‑facing offerings, and delivery methodologies.
• Engage directly with internal stakeholders to translate complex safety and governance requirements into actionable research questions, experiments, and evaluation suites.

Requirements

Education & Track Record

• PhD in Computer Science, Science, Technology and Society Studies (STS), Machine Learning, AI Ethics, Social Data Science or a closely related research domain.
• Proven Research Record: Demonstrated academic excellence with a strong portfolio of peer‑reviewed publications and conference presentations at top‑tier venues (e.g., NeurIPS, ICML, ICLR, ACL, IEEE, CVPR, or dedicated AI Safety conferences).

Technical & Domain Expertise

• Deep Domain Expertise: Substantial experience in AI Safety, Responsible AI (RAI), Trust & Safety, or Model Evaluation.
• Technical Mastery: Strong hands‑on capabilities in Python and modern ML/DL frameworks (e.g., PyTorch, Transformers).
• Generative AI & Agentic Systems: Deep working knowledge of foundation models (LLMs, Text-to‑Image, Multimodal), agents, reasoning systems, and current safety mitigation techniques.
• Benchmarking & Methodology: Proven experience in designing datasets, benchmark suites, red‑teaming protocols, or synthetic simulation platforms.

Mindset & Soft Skills

• Ownership & Self‑Starter Attitude: Proven track record of self‑initiating research projects, bringing in external partnerships, and shaping technical vision.
• Communication & Engagement: Excellent verbal and written communication skills with the ability to convey complex technical concepts to non‑technical stakeholders, executives, and research peers.
• Cross‑Disciplinary Versatility: Ability to operate comfortably across technical data science, social science paradigms, and product execution.

Preferred / Desirable Qualifications

• Prior employment, consulting, advisory, or formal research collaboration with frontier AI labs (e.g., OpenAI, Anthropic, Google DeepMind, Meta AI) or leading safety research collectives (e.g., MATS, GovAI).
• Experience in domain‑specific safety applications, such as election integrity, psychological manipulation, or agentic behavioral safety suites.