Required Qualifications:
• As per the Collective Agreement, must be enrolled as a student in the Winter 2027 semester;
• Practical experience with machine learning in Python, including PyTorch, Hugging Face (Transformers, Datasets) and Jupyter notebooks;
• Familiarity with adversarial machine learning and LLM security concepts, including evasion and poisoning attacks, membership inference, jailbreaking, and direct and indirect prompt injectio;
• Hands-on experience running and evaluating open-weight large language models locally (e.g., Ollama, vLLM), including working on GPU-equipped Linux machines;
• Familiarity with retrieval-augmented generation (RAG) pipelines and LLM agents that use tools (e.g., vector databases, LangChain or similar frameworks);
• Familiarity with the Linux command line, Python virtual environments and basic shell scripting;
• Strong communication skills and the ability to clearly explain technical concepts to students both verbally and in writing, with patience for those new to machine learning.
Preferred Qualifications:
• Experience with LLM red-teaming and evaluation tools (e.g., garak, PyRIT, AgentDojo, HarmBench);
• Experience with privacy-preserving machine learning, including differential privacy and DP-SGD (e.g., Opacus), or federated learning;
• Research experience or graduate coursework in AI/ML security, privacy or trustworthy AI, including reading and critiquing research papers;
• Ability to give feedback on student research projects and writing in a conference paper format (e.g., LaTeX, IEEE/ACM templates).
