cloud & ai summit - ibm
TRANSCRIPT
Cloud & AI SummitExplore solutions to support transformation
The AI Lifecyle and ModelOpsRobin van TilburgAnalytics Architect
2
AI has emerged as a strategic business opportunity
But what if AI goes wrong?
DATA TALENT TRUSTThe lifeblood of AI, but
complexity slows progress
AI skills are rareand in high demand
Skepticism of AI systems & processes
Find it difficult to fundor acquire AI skills
of time spent preparing data versus building AI models
say its very challenging tobuild trust in AI outcomes
Stuck in Experimentation A majority of enterprises face significant challenges in operationalizing AI value
How to operationalize AI?
Based on 2019 Forrester “Challenges That Hold Firms Back From Achieving AI Aspirations”
DevOps: The guiding principles…
Automating the repeatable
Challenge 3: ModelOps requires collaboration…
There are differences between applications and models…
Challenge 1: AI teams will use multiple tools to build AI solutions.
Challenge 2: ML models decay
…between with data science, development, IT, ops, and business teams.
ModelOps is the bridge from model development to model deployment…
ModelOps…
App Deployment
Operational Data
New Requirements & Engagement
DevelopmentTest
Monitoring Retraining
Search for Data
Acquiring Data/ Self Service
Model building
Hadoop
EDW NO SQL
Virtual Data Lake
ModelDeployment
Refining Data
Continuous Delivery of Applications
Continuous Delivery of Models
Infuse AI
Generate Analytical
Data
Collect Data
Organize Data
AnalyzeData
R
Start
DevOps and ModelOps together
AI Lifecycle: Creating and maintaining trusted models
10
Data Exploration
Data Preparation
Model Development
Build
Build
Run
Run
Deployment
RetrainingModel Management
Rebuild models, improve performance and mitigate bias
Monitor and orchestrate models served with WML
Easily deploy models to WML for
online, batch, streaming
deployments
Business KPIs and production
metrics
Fairness & Explainability
Inputs for Continuous Evolution
Infuse
Infuse
Rebuild models, improve performance and mitigate bias
Easily deploy models to WML for
online, batch, streaming
deployments
Monitor and orchestrate models served with WML
AI Lifecycle
11
Data Exploration
Data Preparation
Model Development
Build
Watson Studio
Watson Machine Learning
Run
Deployment
RetrainingModel Management
Rebuild models, improve performance and mitigate bias
Monitor and orchestrate models served with WML
Easily deploy models to WML for
online, batch, streaming
deployments
Business KPIs and production
metrics
Fairness & Explainability
Inputs for Continuous Evolution
Infuse
AI Open Scale
Rebuild models, improve performance and mitigate bias
Easily deploy models to WML for
online, batch, streaming
deployments
Monitor and orchestrate models served with WML
Analyze any data, no matter where it livesConnect to and analyze your data without moving a single byte through dozens of connectors and multiple deployment options
Empower your entire organization with notebooks, visual productivity, and automation toolsLeverage your entire organization with a variety of tools in a single integrated platform
One platform to rule them all from discovery to production Analyze data, build predictive models, and seamlessly integrate Watson Machine Learning to deploy at scale
IBM Watson StudioEnterprise Data Science platform that helps your team work together to build models to make better data driven decisions for your business
13
AI Lifecycle
14
Data Exploration
Data Preparation
Model Development
Build
Watson Studio
Watson Machine Learning
Run
Deployment
RetrainingModel Management
Rebuild models, improve performance and mitigate bias
Monitor and orchestrate models served with WML
Easily deploy models to WML for
online, batch, streaming
deployments
Business KPIs and production
metrics
Fairness & Explainability
Inputs for Continuous Evolution
Infuse
AI Open Scale
Rebuild models, improve performance and mitigate bias
Easily deploy models to WML for
online, batch, streaming
deployments
Monitor and orchestrate models served with WML
IBM Watson Machine LearningEmbed Machine Learning and Deep Learning in your Business
Deploy and Manage ModelsMove models to production, in an easy, secure, and compliant way
Intelligent Model OperationsEmbed intelligent training services, with feedback loops that constantly learn from new data, regardless where it resides
Accelerate Compute Intensive WorkloadsDistribute your deep learning training and Hadoop/Spark workloads with multi-tenant job scheduling
16
AI Lifecycle
17
Data Exploration
Data Preparation
Model Development
Build
Watson Studio
Watson Machine Learning
Run
Deployment
RetrainingModel Management
Rebuild models, improve performance and mitigate bias
Monitor and orchestrate models served with WML
Easily deploy models to WML for
online, batch, streaming
deployments
Business KPIs and production
metrics
Fairness & Explainability
Inputs for Continuous Evolution
Infuse
AI Open Scale
Rebuild models, improve performance and mitigate bias
Easily deploy models to WML for
online, batch, streaming
deployments
Monitor and orchestrate models served with WML
IBM Watson OpenScaleAutomate & Operate AI at Scale
Production monitoring for complianceDetect and mitigate model bias; audit and explain model decisions
Ensure models resiliency in changing situationsDetect drift in data and anomaly in model behavior; specify inputs and triggers to model lifecycle
Align model performance with business outcomesCorrelate model metrics and business KPIs to measure business impact; actionable metrics and alerts
19
Organizations that implement AI get results
35% to 50% 1.5x to 2x 3x and 8x 15% to 30%
Reduced model monitoring effort
Increased models in production
Increased accuracy of models:
1.5x and 2x more AI and machine learning (ML)
models
Forrester New Technology:Total Economic Impact™ Of Explainable AI And Model Monitoring In IBM Cloud Pak For DataStudy Commissioned By IBM August 2020
More more more…. Today!
14:45 - 15:05 A brief History of A.I, from statistical models to Automated Machine Learning-Pt.1
14:20 - 14:40 Discover Cloud Pak for Data
13:25 - 13:45 Watson OpenScale: Manage production AI with trust and confidence in outcomes
Thank you for your attention!Learn more on ibm.com/nl-en/cloud