Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore advanced AI techniques for combating diverse forms of harm on dating platforms at global scale in this 15-minute conference talk. Learn how Tinder addresses harassment, hate speech, pig butchering scams, and underage users through innovative machine learning approaches that leverage pre-trained models and open-source ecosystems. Discover three key strategies for dramatically accelerating modeling pipelines: human-AI hybrid dataset generation for different harm vectors, automated parameter-efficient fine-tuning of open-source large language and multimodal models for violation detection, and efficient real-time serving of fine-tuned adapters using LoRAX and cascade classification. Gain insights into the challenges of training, serving, and maintaining models for the long tail of harmful content while operating at real-time, global scale. Understand how modern AI engineering practices can be applied to trust and safety challenges in consumer applications, presented by a software engineer with experience spanning neuroscience research, financial fraud detection, and AI-powered content generation.
