The revolution of Federated LearningPushed backRamen Dutta argued that raw data access is not necessary for useful analysis because statistical analysis from privacy-preserving tools can provide the needed information.18:10
Guardrails, governance and security in 2021
21 sessions
The revolution of Federated Learning
Private data, Data Science friendlyClaimSynthetic data can be shared more easily, but it can still leak information or nearly copy data from the training set.3:49
PodcastLessons Learned From Hosting the ML Engineered PodcastClaimWorkday's enterprise customers make security and privacy central infrastructure concerns because the company handles human-capital and financial data.34:28
MeetupAgile AI Ethics: Balancing Short Term Value with Long Term Ethical OutcomesClaimPamela Jasper says an audit must examine the full technical and governance life cycle and cannot be fully automated.1:02:40
PodcastThe Godfather Of MLOpsClaimD. Sculley says federated learning is promising for privacy because data can remain on individual devices, but he does not yet know how to make it operational from an MLOps perspective.35:58
MeetupLaw of Diminishing Returns for Running AI Proof-of-ConceptsClaimBefore writing code, teams should ask problem owners about production expectations, budget, maintenance, support, speed, fairness, regulation, and privacy.21:01
PodcastModel Performance Monitoring and Why You Need it YesterdayClaimAmit Paka said Fiddler AI's mission is to help teams build trust with AI.8:02
PodcastMaturing Machine Learning in EnterpriseClaimKyle Gallatin says machine learning has moved from isolated proof-of-concept data science projects toward MLOps, and that governance, observability, and visibility are the next stage.10:13
MeetupEngineering MLOpsClaimPoor model release strategies, weak governance, insufficient auditing, and inadequate real-world testing are additional causes of failure.8:21
PodcastEnterprise Security and Governance MLOpsPushed backDiego Oppenheimer argues that security investment should be a conscious risk-reward decision rather than an identical day-one requirement for every organization.35:09
MeetupMachine Learning in CybersecurityPushed backMonika argued that machine learning cannot completely replace cybersecurity experts, while Demetrios Brinkmann questioned whether new attacks would be missed by the models.18:21
PodcastCreating MLOps StandardsClaimAlex Chung says enterprise teams need flexibility to choose tools while maintaining common interfaces, governance, spend management, and resource allocation.12:31
PodcastML Security: Why should you care?ClaimSahbi Chaieb says that fetching external data or models from the internet creates a security risk because they may be malicious or contain threats.6:13
MeetupBuilding Machine Learning Models into Docker ImagesPushed backLuke Marsden argued that building containers inside containers with a mounted Docker socket is a security problem.49:21
MeetupEngineering Best Practices for Machine LearningClaimThe engineering practice catalog groups practices into data, training, coding, deployment, teams, and governance.8:48
PodcastThe Future of Data Science Platforms is AccessibilityClaimSkylar Payne says privacy practices should be person-centered, clearly explain why data is collected, give people control, and support graceful degradation when people do not share everything.31:37
Podcast2022 Predictions for MLOps and the IndustryPushed backDemetrios Brinkmann raised EU regulation as a robustness issue, while Reah Miyara said he was less familiar with the laws and governance involved and that defining robustness would be difficult.32:36