This is an in-person event! Registration using this form is required in order to get in. Github will email you a form the day before the event, which you will need to complete for your access pass.

Topic: Connecting your unstructured data with Generative AI

What we’ll do:
Have some food and refreshments. Hear three exciting talks about unstructured data and generative AI.

5:30 - 6:30 - Welcome/Networking/Registration
6:35 - 7:00 - Yury Malkov, Research Scientist, OpenAI
7:05 - 7:30 - Jithin James, CEO and Shahul ES, Co-Founder, Ragas
7:35 - 8:00 - Jason Lopatecki, CEO, Arize
8:00 - 8:30 - Networking

Who Should attend:
Anyone interested in talking and learning about Unstructured Data and Generative AI Apps.

When:
February 20, 2023
5:30PM

Tech Talk 1: Approximate Nearest Neighbor Search in Recommender Systems
Speaker: Yury Malkov
Abstract: I am going to discuss problems and research regarding Approximate Nearest Neighbor Search in Recommender Systems. In particular, the role of fast Approximate Nearest Neighbor (ANN) search in the multi-stage funnel design or a typical Recommender System. I'll discuss research on ANN search with neural ranking distances and its impact on the end-to-end funnel design.

Tech Talk 2: Metrics Driven Development of RAGs
Speaker: Jithin James and Shahul ES
Abstract: We will be walking through a RAG application from scratch with a metrics diven approach. We'll startoff with a very basic RAG system, figure out the most gaps in performance and make improvements along the way guided by proper evaluations of the RAG pipeline.

Tech Talk 3: Path to Production: LLM System Evaluations and Observability
Speaker: Jason Lopatecki
Abstract: As new LLMOps tools race to keep up with the latest capabilities of foundation models, generative AI is at a crossroads. Over half (53.3%) of machine learning teams are planning production deployments of LLMs in the next year, but many continue to cite issues like hallucinations and responsible deployment as barriers in moving LLM-powered systems into the real world.
In evaluating LLM-powered apps, human feedback is paramount – but in practice is not available for most sub-calls. This talk by Arize CEO and co-founder, Jason Lopatecki, covers how teams can achieve fast and accurate LLM-assisted evaluations and apply data science rigor to the testing of model and template combinations. In addition to covering LLM evals, the talk will dissect open-source tooling and soon-to-publish results on observability best practices learned from the trenches.

Where:
This is an in-person event. Registration using this form is required to get into the event. Registration in advance will close 2 days before the event. Co-sponsored by Zilliz maintainers of Milvus and Arize.