AI Red Team Expert (Freelance) - Protect the Future of Language AI
LILT is revolutionizing global communication with AI. Join us! We're seeking talented and passionate freelance AI Red Team experts to help us secure the next generation of AI-powered language solutions. If you live and breathe cybersecurity and have a knack for finding vulnerabilities, this is your chance to make a real impact.
As a key member of our Red Team, you'll be on the front lines of adversarial testing, challenging the security and robustness of our AI systems – including LLMs, multimodal models, inference services, RAG/embeddings, and product integrations. Think like an attacker, find the weaknesses, and help us build stronger, more resilient AI.
What You'll Do:
Craft creative prompts and scenarios to rigorously test model guardrails.
Explore innovative methods to bypass restrictions and uncover vulnerabilities.
Systematically document testing outcomes and provide clear, actionable insights.
Collaborate closely with engineers and safety researchers to share findings and improve system defenses.
What We're Looking For:
We need someone who thinks like a hacker but acts like a guardian. Here's what we're looking for:
Key Technical Skills:
Generative AI Expertise: Deep understanding of Generative AI models, including LLMs, their architectures, training processes (e.g., prompt engineering, fine-tuning, RLHF), and potential failure modes.
Cybersecurity & Threat Modeling: Solid experience in cybersecurity principles, threat modeling, vulnerability assessment, and penetration testing. Ability to identify attack vectors, simulate real-world threats, and understand their potential impact.
Data Analysis & NLP: Strong analytical skills to dissect model outputs, identify biases or errors, and recognize patterns in model behavior. A background in Natural Language Processing (NLP) is a major plus.
Ethical Hacking Mindset: A strong commitment to using your skills for defensive security, adhering to a strict ethical code, and understanding responsible disclosure.
Core