Red Team LLM Security Researchers

Job Description

Role & responsibilities We are looking for highly skilled Red Team LLM Security Researchers to test, evaluate, and enhance the robustness of AI language models. This role involves crafting adversarial prompts, identifying vulnerabilities, and working with AI engineers to improve model security and reliability. Design and execute adversarial attacks against LLMs to identify vulnerabilities in hallucination, bias, and security. Develop sophisticated prompt engineering techniques to induce and analyze model failures. Collaborate with AI engineers, linguists, and cybersecurity experts to strengthen model defenses. Test for model exploits such as prompt injection, misinformation propagation, and ethical loopholes. Document vulnerabilities and work with technical teams to implement mitigations. Research and stay ahead of emerging adversarial AI threats and LLM red teaming methodologies. Assist in developing automated testing frameworks for adversarial robustness. Create adversarial image and text instructions to bypass alignment and then probe the model with attack prompts, recording outputs and focusing on cases where the model fails or struggles. Classify and group all successful prompt attacks by attack technique. Score responses based on accuracy, tone and clarity, and safety and ethical considerations. Validate attack classification with the model, logging incorrect classifications. Add feedback where responses can be improved (e.g., adding specific examples, citations to reputable sources). Probe detector performance with model outputs, logging failed detections (e.g., model response is not factual). For failed faithfulness/groundedness responses, outputs will be rewritten for factual correctness based on inputs from previous model turns Qualifications Experience in red teaming projects. Education in one of the following areas: Computer Science: Offers a broad understanding of algorithms, programming, and systems, crucial for developing and testing AI models. Cybersecurity: Specializing in cybersecurity provides specific skills in identifying and mitigating vulnerabilities, essential for red teaming. Artificial Intelligence/Machine Learning: Provides in-depth knowledge of how these models work, vital for understanding their potential weaknesses. Data Science: Combines statistical analysis, machine learning, and data management, important for working with AI models. Ethical Hacking/Penetration Testing: Some programs specifically focus on ethical hacking and penetration testing, which are directly applicable to red teaming Contact Send ur resume careers@eliscription.com Ravi Teja HR 9866005442 / Whats App Also Address: Elico Healthcare Services Ltd B-91,APIE, Gate No. 6, Sanath Nagar,Hyd Near Bharat nagar MMTS Railway station. Regards, Team HR