The Mythical LLM-Month: Building Trustworthy Copilots and AutoPilots
Claudionor N. Coelho Jr, PhD/MBA
Chief AI Officer, Zscaler Inc
Date: Friday, May 31
Time: 1:00 pm PT
YouTube link: https://youtu.be/S8cop6_wDqQ
Abstract
No doubt. Last year was the year of the LLM. As we move more and more into 2024, we are seeing that almost no LLM promised last year was delivered, or when delivered, they quickly made the headlines.
We are going to show that what people have been claiming to be standalone LLMs are in fact complex software systems where LLMs are an integral part of it. We call these systems AI Agents. A lot of solutions built are still in their infancy, and the race to get these systems up and running caused a huge impact in readiness of such systems. For example, one important aspect of AI Agents to increase their reliability is grounding, which is often forgotten in examples and demos. Finally, most people still regard data as a second class citizen, and in reality for AI Agents based on Retrieval Augmented Generation (or RAG for short), having good data is as important or more important than a LLM.
We will finalize this presentation showing how to merge software engineering practices from the last 50 years into this new era, where multimodal data and algorithms are integrated into vector databases, and used to create high quality copilot applications.
About
With nearly three decades in the AI, software and semiconductor industry, Claudionor has extensive research and development experience in machine learning and deep learning techniques (including Generative AI), software systems, cybersecurity and semiconductors. As the Chief AI Officer at Zscaler, he is chartered with driving the vision, strategy and implementation of advanced AI technologies to strengthen the world’s largest security platform and propel Zscaler’s innovation engine forward.
Prior to joining Zscaler, Claudionor served as the Chief AI Officer and SVP of Engineering at Advantest, where he spearheaded the development of a Zero Trust private cloud solution tailored for the semiconductor manufacturing market, enabling it to run Machine Learning workloads. Before that, Claudionor was the VP/Fellow of AI and the Head of AI Labs at Palo Alto Networks where he led the charge in AI, AIOps and Neuro-symbolic AI, an advanced form of AI that enables reasoning, learning, and cognitive modeling, to help revolutionize time series analysis tools on a massive scale. Claudionor’s career also includes vital roles in Machine Learning and Deep Learning at Google, where he developed a state-of-the-art Deep Learning technology designed for automatic quantization and model compression which played a pivotal function in the search for subatomic particles at CERN. This work was featured in the cover page of Nature Machine Intelligence in August, 2021. He has also served as a senior executive in several EDA companies, namely Synopsys, Cadence, Jasper Design Automation and Verplex Systems.
Claudionor holds a PhD in Electrical Engineering and Computer Science from Stanford University and an MBA from Ibmec in Brazil. He earned both his M.Sc in Computer Science and B.S. in Electrical Engineering from the Universidade Federal de Minas Gerais and is an Invited Professor in the Electrical and Computer Engineering Department at Santa Clara University.