Your personal research assistant. coming soon…

Background I’ve spent the last few months excitedly building products. Recent advancements in deep learning, and Gen-AI in particular, have given me a new optimism for what’s possible and how quickly and effectively it can be built. It’s been exhilarating to build quick prototypes on top of foundation models and get magical results. However, its surprisingly difficult to get from a interesting prototype to genuinely delightful product. It seems like the rigour that’s traditionally associated with applying a deep learning model is nascent when building products with Gen-AI. So I was excited to come across DSPy, it’s “programming over prompting” approach, and the potential to make a more robust product experience on top of Gen-AI. In this post I’ll walk through how I used DSPy to rebuild the event ingestion part of an old project. ...