MeetupHierarchy of Machine Learning NeedsThreads / The platform team / 2020
The platform team in 2020
21 sessions
MeetupBuilding an ML Platform at SurveyMonkeyClaimSurveyMonkey chose to build its platform because no single product met its data, serving, and management requirements, and the company had relevant expertise in-house.5:50
MeetupMLOps - The Blind Men and the ElephantClaimMachine learning engineers are usually strong software engineers and machine learning practitioners who were already adopting MLOps practices before the term became common.6:46
MeetupMachine Learning at Scale in Mercado LibreClaimBefore the shared platform, Mercado Libre teams used different technologies and approaches, including IBM Watson, custom Python pipelines, and other services.9:47
MeetupFury Platform and Fury Data Apps at Mercado LibreClaimThe platform team is trying to enable external innovation by integrating services built and maintained by other teams, including a feature catalog for fraud-related entities and representations.9:40
MeetupRunning a Fintech on Machine LearningPushed backThe team favored building its feature store internally rather than buying an existing product because it wanted tighter integration and customization, and found no sufficiently mature market solution.46:45
MeetupBuild vs Buy an ML PlatformPushed backDiego Oppenheimer rejects the assumption that buying a platform necessarily removes flexibility, comparing it with the mix of purchased tools used in software development.40:57
MeetupHow to Leverage ML Tooling EcosystemClaimMariya Davydova says Neu.ro is a resource-orchestration platform built on Kubernetes and designed to work across cloud, on-premises, and hybrid infrastructure.8:59
PodcastMLOps: Isn't That Just DevOps?ClaimTeams should first understand their particular use case and requirements before choosing tools or deciding whether to build or buy an MLOps platform.55:35
MeetupScaling ML Capabilities in Large OrganizationsPushed backJoe argues that organizations should avoid buying a model serving platform when a simpler, opinionated solution is sufficient.57:25
PodcastHow to Choose the Right ML ToolPushed backMariya Davydova recommends building an internal MVP in the Feast situation, while the team is also considering waiting for the planned release or choosing another tool.33:31
MeetupUN Global PlatformPushed backMark Craddock argued that a platform should use cloud infrastructure rather than building equivalent capabilities in its own data center.25:27
The intersection between DataOps and privacyClaimA privacy-compliant data platform must keep personal information organized and easy to locate so that it can be deleted or extracted when required.22:36
MeetupHuman-centric ML Infrastructure: A Netflix OriginalClaimMetaflow is Netflix's machine learning framework and is used internally for use cases including content analysis, fraud detection, and intelligent infrastructure.2:38
MeetupHow To Move From Barely Doing BI to Doing AIClaimJoe Reis says cloud platforms make it increasingly easy to build machine-learning pipelines, especially for cloud-native companies using services such as SageMaker and Athena.14:03
PodcastSRE for ML InfraClaimTodd Underwood says clearly defined interfaces and APIs allow users to replace individual pipeline components without depending on how those components are implemented internally.21:25