ML for Systems and Systems for ML
summary
Frames the field as two concrete directions: building systems to run ML and using ML to improve systems. Lists actual platforms and results: TensorFlow and Torch and hardware like GPUs and TPUs power scalable training, plus PG Vector for vector search inside Postgres. On the reverse side, ML models are applied to real system problems: Ulus and Bako use learned policies to optimize data‑plane routing and compiler flag selection, and DB tune applies ML to automate Postgres configuration, replacing manual heuristics with adaptive learned policies.
AI-written summary · about
related talks
54mML Beyond Metrics: Challenges of Using ML Systems in the Real World
26mMLbase: A Distributed Machine Learning System
43mML application lifecycle
55mML with real impact driven by OS technologies
26mPatterns of ML Models in Production
41mInfrastructure for ML Applications
33mBuilding Resilient (ML) Pipelines for MLOps
31mUse-Case Driven ML Platforms