Explore advanced techniques for accelerating Bayesian inference and experimental design through amortization in this 41-minute seminar presentation. Learn how neural networks can be pre-trained with synthetic data to dramatically reduce computational costs during deployment, addressing the notorious computational challenges inherent in traditional Bayesian methods. Discover the Amortized Conditioning Engine (ACE), a flexible framework that enables conditioning on both observed data and interpretable latent variables while incorporating priors at runtime and generating predictive distributions for discrete and continuous data. Examine the cutting-edge Amortized active Learning and INference Engine (ALINE), which unifies amortized inference and experimental design into a single framework capable of rapidly proposing valuable data points while performing fast, flexible inference on collected data. Understand how these approaches create a seamless loop between active data acquisition and real-time reasoning, making them particularly valuable for critical applications requiring strategic data acquisition and instantaneous inference under uncertainty.