Under the Hood - How Architectures Shape What Models Can Do

Generative AI is transforming the way we build intelligent systems — but not all models are created equal. In this talk, we'll demystify the major architectures behind Gen AI models, including encoders (like BERT), decoders (like GPT), and encoder-decoder hybrids (like T5 and FLAN). We'll explore how these architectures differ in design and capability, and provide practical guidance on which to use — and why — depending on your use case, whether it's classification, generation, retrieval, or fine-tuning. This session is ideal for data engineers, ML practitioners, and product leaders looking to move beyond the hype and toward solutions that scale.