Launch Event: Lexsi AI Alignment and Safety Hackathon

Join the Lexsi Labs team at Polaris School of Technology, Bengaluru, for an afternoon on AI safety and post-training, with speakers from industry and academia and a curated audience working on how AI models are refined after initial training and how to make their behavior safer and more reliable. We'll explore: Fine-tuning and Data Curation: curating and generating training data, post-training models, and evaluating and benchmarking the resultsAI Red Teaming, Safety and Alignment: auditing and repairing safety drift in fine-tuned models, and locating and intervening on the circuits behind itAgentic Fine-tuning: post-training agents for better tool calls, syntax and efficiency, with fewer redundant or incorrect calls Hackathon kickoff: We will discuss about the agenda, goal and the participation criteria for the biggest hackathon on AI Safety and Alignment. Registration for the Lexsi Labs Hackathon with Cohere Labs and Hugging Face is open with over 30000$ in prizes and credits. At the kickoff, we'll walk through the three tracks and the open-source tools they build on, followed by sample problem statements. Q&A and networking: Stay after to put your questions to the speakers and the Lexsi Labs team, swap perspectives, and meet potential collaborators and teammates. We're hiring: Lexsi Labs is hiring across research, engineering and developer relations across levels (US, PG, PhD). If you're working on any of this, come talk to us. Agenda: TBA Speakers: TBA