[Intern] AI Applied Researcher Internship

Join the Accenture Labs’ AI R&D Team and work on core AI Models / technology that work across multiple domains The Opportunity As an intern on our team, you will work at the frontier of next-generation AI systems to explore how they can become more reliable. Large Language Models (LLMs) are powerful but overconfident — they often produce fluent, plausible, but incorrect answers. This project aims to fine-tune a compact open-source LLM (like Qwen-2-7B, Mistral-7B, or DeepSeek-6.7B) so that it can quantify its own uncertainty and explicitly say “I don’t know” when faced with unknown or ambiguous questions. The main effort is to train the model using Evidential Deep Learning (EDL) — a technique that replaces the standard softmax-based confidence mechanism with a Dirichlet-based evidence distribution. This allows the model to represent not just what it predicts, but how certain it is. The end goal is a “humble” AI assistant: one that maintains accuracy on known information but signals epistemic uncertainty clearly, improving reliability, interpretability, and safety. Key Responsibilities & Impact This role offers a unique mix of deep research and hands-on development: Advanced Research Synthesis: Review, analyze, and strategically summarize the latest academic and industry research in technical domains (e.g., uncertainty quantification in LLMs). Hands-on Prototype Development: Collaborate directly with Senior Research Scientists to rapidly design, build, and test PoCs. Knowledge Mobilization: Articulate complex research findings into clear, persuasive, and publishable papers and presentations for both academic and business audiences. Strategic Storytelling: Craft engaging narratives and technical demonstrations that translate research outcomes into clear, compelling value propositions.