Slides on retrieval and ranking with ONNX inference

The wonderful DataPhilly meetup community asked me to come give a talk on any topic, and I talked for half an hour on embedding-based retrieval, machine learning model artifacts creation and inference, general versus smaller models, and understanding latency and ONNX conversion tradeoffs. People also had some fantastic questions around what can we embed, tradeoffs in hosting your own models versus API for inference, and whether a model’s data or achitecture is more important.

It was especially meaningful for me to give this talk at DataPhilly, my second time speaking there. The first time, I gave a talk in 2013, I was a junior data analyst, and my talk was about the frustrations of data access. It turns out that the way to solve those frustrations was for me to spend the next decade plus becoming a developer and writing some gnarly access patterns of my own. Good times!

Slides are below. You can also download the PDF directly.