REX – GenAI in prod : practical tech & product insight for delivering value with RAG

Synopsis

GenAI and especially large language models (LLMs) are reshaping the tech and AI ecosystem since ChatGPT release in nov 2022. Since then, an increasing number of organisations are sharing the weights of their pretrained and finetuned model with relatively permissive license. Among the use cases of LLM, the possibility to use them on top of a database of (un)structured documents to perform retrieval and synthesis for question answering with accurate semantics is booming – this is Retrieval Augmented Generation (RAG).

If it sounds promising, building a RAG-based solution that actually solves a user’s need is no easy task. On the tech side, choosing the appropriate models for embedding and generation, evaluating performances, and monitoring costs have no silver bullet solutions as of today and many machine learning engineer teams develop their own pipeline. Meanwhile, on the product side, articulating the unique value proposition of RAG and educating the market about its potential benefits and limitations remains a challenging task when handling data protection and dealing with hallucinations. This meetup is a synthesis of interviewing several organisations, with various level of maturity in their LLM and RAG journey to gather insights and share with you actionnable tips you can leverage in building incredibly helpful products.

Watch the replay

Full recording of the talk at DataCraft Paris, REX – GenAI in prod : practical tech & product insight for delivering value with RAG

Audience

CTO, Machine learning engineers, AI product owner, data scientist

Key Takeaways

Technical Challenges of RAG: we’ll review the complexities of selecting models, evaluating performance, and managing costs in the development of RAG-based solutions.
Product Strategy: we’ll share insights into defining the unique value proposition of RAG for your users and navigating the market’s learning curve.
Actionable Tips: we’ll distill the feedback of interviews with several C levels, tech lead and ML lead from various companies on their experiences in RAG implementation.