Explore the next frontier of machine learning systems in this 15-minute conference talk from Conf42 Observability 2024. Delve into the concept of model observability, its importance, and how it differs from model monitoring. Discover key components and challenges in ML observability, including data drift, performance degradation, and data quality issues. Examine specific challenges and evaluation techniques for Large Language Models (LLMs) and computer vision systems. Learn about monitoring techniques, explainability methods for various ML systems, and gain insights into future trends in model observability. Enhance your understanding of this crucial aspect of ML system development and maintenance.
Overview
Syllabus
intro
preamble
who am i?
what is model observability?
how is model observability different from model monitoring?
why is model observability important?
how ml observability helps
key components of ml observability
key challenges: data drift
key challenges: performance degradation
key challenges: data quality
model observability challenges in llms
evaluation techniques for llms
challenges in computer vision
components to address challenges in computer vision
monitoring techniques in ml observability
explainability techniques in standard ml systems
explainability techniques in llms
explainability techniques in cv
future trends in model observability
thanks for attention
Taught by
Conf42
