Description
We are looking for a Senior Offensive AI Security Researcher with a strong offensive security and research mindset to lead novel adversarial research across Salesforce’s AI ecosystem - uncovering critical and systemic vulnerabilities before they can be exploited.
This role goes beyond finding individual bugs - it focuses on scoping, designing, and executing advanced research into emerging AI threat vectors, combining academic research with practical offensive methodologies to establish new TTPs targeting complex AI paradigms, including indirect prompt injection, tool abuse/excessive agency, RAG system poisoning, data extraction, and model/data supply chain compromises.
You will work as a trusted partner to engineering and architecture teams - providing an offensive security perspective that shapes internal threat models, guides AI architecture decisions, supports corporate governance, and directly contributes to Salesforce’s Generative AI Security Standard.
Responsibilities
- Lead Novel AI Security Research: Scope, design, and execute advanced adversarial research across Salesforce’s AI ecosystem, applying a risk-based approach to uncover novel threat vectors before exploitation.
- Innovate & Operationalize AI Attack Techniques: Combine academic research with practical offensive methodologies to establish new TTPs targeting complex AI paradigms- specifically indirect prompt injection, tool abuse/excessive agency, RAG system poisoning, data extraction, and model/data supply chain compromises.
- Guide Strategy & Security Standards: Provide an offensive security perspective to guide AI architecture, shape internal threat models, support corporate governance, and directly contribute to Salesforce’s Generative AI Security Standard.
- Disseminate Research: Publish internal and external threat intelligence, vulnerability analyses, and security research to keep Salesforce defenses aligned with state-of-the-art AI threat landscapes.
Requirements
- 5+ years of experience in offensive security roles (vulnerability research, penetration testing, application security, or security engineering).
- 2+ years of hands-on experience auditing, breaking, or researching ML/AI systems, with deep expertise in LLM attack surfaces and frameworks (e.g., OWASP Top 10 for LLMs, MITRE ATLAS).
- Deep technical understanding of modern AI design patterns, including Retrieval-Augmented Generation (RAG), autonomous agentic workflows, model context protocols, and AI pipelines.
- High degree of proficiency in Python for exploit proof-of-concept development, data analysis, and technical evaluation.
- Proven track record of leading complex security initiatives and communicating critical, high-stakes technical risks to both research engineers and executive leadership.
Nice to Have
- Advanced degree (MS or PhD) in Computer Science, Machine Learning, Cybersecurity, or a related field.
- A public portfolio of offensive security research, including conference presentations (e.g., Black Hat, DEF CON, NeurIPS), published papers, CVEs, advisories, or open-source research contributions.
- Experience auditing modern AI development ecosystems and tools (e.g., PyTorch, Hugging Face, LangChain, vector databases, or open-weight models).