Generative AI: Best practices of embedding models and LLM projects


Details
** PLEASE NOTE THAT THE MEETUP IS AT 3 pm KSA // 4pm Dubai Time. DO NOT BE CONFUSED BY THE "AST" timezone displayed by the website **
**
Zoom link: https://mckinsey.zoom.us/j/99201845814?pwd=cDVkdHA1eUxibWlYU2tuSXlIM0lOdz09
**
Details
We will be hosting this QuantumBlack Middle East Meetup virtually on Thursday, 23rd November 2023, 3 pm KSA / 4 pm DBI / 1 pm CET
The session will focus on Generative AI: Best practices of embedding models and LLM projects
***
AGENDA (AST, GMT+3):
15:00 PM – 15:10 PM: Zoom Link Active, Welcome & Introductions
15:10 PM – 15:30 PM: "Deep dive into embedding models" by Pascal Brokmeier, Expert at Quantumblack, AI by McKinsey, Amsterdam
15:30 PM – 15:40 PM: Q&A
15:40 PM – 16:00 PM: "How not to fail a Data Science project, involving Large Language Models**”** by Alexey Miasnikov, AI expert and Data Science leader, Careem
16:00 PM – 16:10 PM: Q&A
16:10 PM – 16:20 PM: Closing
***
Zoom link: We will update the link here closer to the date
Speakers:
Session 1: Deep dive into embedding models by Pascal Brokmeier
Given the “data within the region” policy and the limited data centers by the cloud providers, we cover the different options of how to pre-process the data to be self-hosted via embeddings model, the different options to host vector databases as well as key criteria’s to select the right vector database.
Bio:
Pascal Brokmeier is a Data Engineer & Cloud Architect with 10+ years of experience in data systems architecture & engineering as well as digital product leadership. At QuantumBlack, he leads the GenAI research and focusses within the Life Science Practice. He is currently leading the development of a platform that allows integrating GenAI tools and services across the entire portfolio of QuantumBlack products and solutions.
Session2: How not to fail a Data Science project, involving Large Language Models by Alexey Miasnikov
What is beyond the LLM capacity and which expectations are reasonable? What do you need under your belt and technology stack to tame LLMs and avoid "hallucinations" in their responses? How to eliminate biases in LLM responses? Reasonable rationales to tune LLMs and outcomes of fine-tuning. All these and other questions will have solid answers during the session, backed by comprehensive review of fresh literature and practical recommendations.
Bio:
PhD, Data Science manager with 15 years of experience currently at Careem, ex-Tokopedia, ex-Booking.com. Built large-scale recommender systems, massive fraud prevention solutions, applied LLMs to model millions of e-commerce customers, and many others via thought leadership and empowered teams
Sponsors:
This event is sponsored by QuantumBlack, AI by McKinsey.
QuantumBlack is an advanced analytics firm operating at the intersection of strategy, technology and design to improve performance outcomes for organisations.
***
Notice:
By participating in the meetup you agree that McKinsey & Company may (i) videotape, audiotape, photograph, or otherwise record your name, voice, or image, and (ii) use and distribute such videotapes, audiotapes, photographs or recordings of your name, voice or image in any texts, videos, and other materials that McKinsey may make available through websites and social media to its employees or any third parties.
If you do not wish to be recorded please inform the event organiser in advance.

Generative AI: Best practices of embedding models and LLM projects