BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium - Finally: Autonomous Systems
DTSTART:20251117T161500
DTEND:20251117T171500
DTSTAMP:20260415T183658Z
UID:1c13b5c8b461efc544a5f6ac5bf2e7b54fc5758de6782375bb4f48d0
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: JP Vasseur - NVIDIA\nVideo of his talk\n\nAbstract\nFor 
 two decades\, the promise of truly autonomous systems—characterized by s
 elf-learning\, self-healing\, and self-governance—has remained a persist
 ent\, yet largely unrealized\, ambition in computer science. We posit that
  this inflection point has now arrived. The synergistic convergence of dee
 p statistical analysis\, predictive machine learning\, and sophisticated a
 gentic frameworks powered by generative AI provides\, for the first time\,
  the technological foundation to achieve this long-sought autonomy.\n\nThi
 s presentation introduces a novel architectural framework for a truly auto
 nomous\, self-improving system designed for the operational management of 
 large-scale\, complex distributed environments. We propose a cognitive sys
 tem built upon a fleet of collaborative\, specialized AI agents. This mult
 i-agent architecture integrates heterogeneous models—spanning statistica
 l machine learning for predictive analytics and generative AI for complex 
 reasoning and knowledge synthesis. We will deconstruct the core principles
  of this approach\, addressing key scientific challenges inherent in build
 ing such autonomous systems.\n\nCore topics will include: managing system 
 stochasticity\; ensuring the interpretability of agentic decision-making\;
  the efficacy of knowledge compression techniques for interfacing with Lar
 ge Language Models (LLMs)\; and methodologies for measuring operational ef
 ficacy in a closed-loop system. We will demonstrate why a proactive\, pred
 ictive posture is a critical for autonomy\, exploring the role of machine 
 learning in identifying failure patterns before they manifest. Furthermore
 \, we will analyze the challenges of emergent reasoning trajectories withi
 n the multi-agent system and outline a framework for enabling continuous s
 elf-improvement. This work presents a conceptual and practical blueprint f
 or the next frontier of autonomous systems\, moving beyond mere automation
  to achieve genuine operational self-governance.\n\nBio\nJP Vasseur is a 
 pioneering technologist whose innovations have shaped Networking\, Artific
 ial Intelligence\, and large-scale compute systems for over three decades.
  A key contributor to Networking( PCE\, MPLS\, and Traffic Engineering)\, 
 he has also led advancements in Machine Learning and Generative AI\, driv
 ing the convergence of AI and infrastructure in the Internet (WAN\, Wifi\,
  DC). As Senior Distinguished Engineer and Chief Architect\, AI & Networki
 ng at NVIDIA\, JP leads the design of agentic\, self-operating data center
 s capable of autonomous issue detection\, root-cause analysis\, and remed
 iation. Previously\, during more than 25 years at Cisco\, including 13 as 
 Fellow and AI VP of Engineering\, he pioneered AI-driven solutions and led
  team shipping products deployed at large scale. (in IoT\, SD-WAN\, Wifi a
 nd Security)\n\nHolder of over 750 patents and a PhD in Computer Science\
 , JP is a globally recognized inventor\, author\, and thought leader advan
 cing the intersection of AI\, networking\, and neuroscience.\n\nMore infor
 mation
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420 https://epfl.zoom.us/
 j/63795601113
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR