From b230dd58d174420d94735364e6f28cd1b29c4264 Mon Sep 17 00:00:00 2001 From: Pascal Zaragoza Date: Tue, 2 Apr 2024 14:50:02 +0000 Subject: [PATCH] submission for talk about LLM use in Pharo --- .../talks/2024-LLM-RAG-Pharo.pillar | 26 +++++++++++++++++++ 1 file changed, 26 insertions(+) create mode 100644 2024-Conference/talks/2024-LLM-RAG-Pharo.pillar diff --git a/2024-Conference/talks/2024-LLM-RAG-Pharo.pillar b/2024-Conference/talks/2024-LLM-RAG-Pharo.pillar new file mode 100644 index 00000000..62ae6ec7 --- /dev/null +++ b/2024-Conference/talks/2024-LLM-RAG-Pharo.pillar @@ -0,0 +1,26 @@ +""Title"": +How to get more people onboard with Pharo ? Applying LLM as support for the onboarding of new developers. + +""Name"": +Pascal Zaragoza, Marius Pingaud + +""Email"": +pascal.zaragoza@berger-levrault.com, marius.pingaud@berger-levrault.com + +""Abstract"": +Last year, Large Language Model (LLM) have taken the spotlight in many fields with their ability to generate text in an intelligent manner. +In computer science, they are being used to generate code, documentation, and also to explain any code base. +These use cases rely on a technique called Retrieval Augmented Generation (RAG). +This technique relies on retrieving the relevant document or code just in time to perform the task. +In this work, we explore RAG to study their use in the onboarding of developers toward new technologies. +As beginner in Pharo ourselves, we have developped a RAG on the Pharo code base and tested it on a set of questions from the official Pharo Discord where many new developers go to seek answer from the community. +In this talk, we present a quick overview of LLM and how to build a RAG on any given documents (e.g., Pharo source code and public books of Pharo). +Then, we demonstrate the ability of RAG to automatically help onboarding pharo users, based on previous questions found in the Pharo Discord. + +""Bio"": +Pascal Zaragoza has a PhD in computer science from the university of Montpellier. +His thesis was on the migration of monolithic architecture toward microservice using model driven engineering. +Since 2022, he works on integrating AI/LLM inside the workflow of Berger-Levrault products. +Marius Pingaud is a computer science student in this first year of master at the university of Sorbonne Paris. +He is also a part time intern at Berger-Levrault, under the supervision of Phd Pascal Zaragoza. +His work focuses on the development of a RAG platform for Berger-Levrault and he explores the RAG ability to accelerate the onboarding of new developers. \ No newline at end of file