Building AI applications is a very complex process involving steps and workflows which are becoming more complex every other day. Its a circle since the AI application is nothing but a feedback loop between various steps involving data. Consider the below picture a data scientist or ML engineer has to work through. Now my mission as an evangelist of the AI technology who sees a lot of promise in this technology would like to make it simple so we can empower more professionals in the business to become what we call "citizen data scientists". A citizen data scientist is a business person empowered so well that he can combine his domain knowledge with tools an expert data scientist uses in a simplified way. We have seen this impacting customer experience in 5x and revenue increase in the range of 15-20%.
1. AI for AI
Karan Sachdeva
Director Cloud Pak for Data,
Asia Pacific
IBM Data & AI
karan@sg.ibm.com
2. Use Cases “86% of CEOs consider advanced analytics(Artificial Intelligence & Data Science) to be
the priority № 1 for their companies”
Source PwC General Directors Survey Companies
8. AutoAI: What we are hearing from our clients
Healthcare
insurer
AutoAI is my first step now with any new
use case, I can test it out and see
immediately if there is signal in the data
Airline Leader
Demonstrated with live data that AutoAI
outperforms the existing loyalty
campaigns designed by their in-house
marketing teams.
International Banking
Leader
The AutoAI function has played an
important role in the feature engineering
process. We are super excited and want
to get deeper dive on the potential of
AutoAI
9. Why care about Auto AI?
“Through 2020, the
number of citizen data
scientists will grow
five times faster than
the number of expert
data scientists.”
Source: Gartner: Top 10 strategic
technology trends for 2019,
October 2018
Empower
Citizen Data Scientists
Speed
AI lifecycle
management
“The moment you
put a model in
production, it
starts degrading.”
Source: Forbes, Why Machine
Learning Models Crash And Burn
In Production, April. 2019
“Typical project
consumes 1-6
expert data
scientists for 2-12
months.”
Source: IBM Research estimate,
March 2019
Enhance
Performance &
Reproducibility
10. Domain
Knowledge
Auto AI
Automated Machine
Learning and Data
Engineering
Data
Science
Math and
Statistics
Computer
Science
and Coding
Machine
Learning
Data
Science
Application
Software
Industry
Applications
Domain
Knowledge
The Case for Auto AI
11. AI for AI: Automation of AI Development with Cloud Pak for
Data
AI Designing AI AI Optimizing AI AI Governing AI
Create AI and Neural
network Models
AI Model optimization Monitoring AI outcome for
explainability
12. 12
AI Designing AI: AI Workflow’s Bigger & More Complex
Majority spent on
data wrangling!
Use Case
Business Req
Data Cleansing
Feature
Engineering
Model Selection
Parameter
Optimization
Ensemble
Model
Validation
Model
Deployment
Runtime
Monitoring
Model
Improvement
Source: https://www.kaggle.com/paultimothymooney/2018-kaggle-machine-learning-data-science-survey
AI Lifecycle
Management
13. AI Optimizing AI
Raw Labeled
Data Set
Prep
Model
selection HPO
Feature
Engineering
HPO
Finds best
preprocessing
imputation /
encoding and
scaling strategies
Finds top-K
estimators
HPO on
selected
estimator
HPO on
estimator
after Feature
Engineering
Finds best data
transformation
sequence
14. “AI Governing AI”: Trust,
and Transparency with
Watson OpenScale via Cloud
Pak for Data
Understand where data originated
(data provenance)
Understand how models are making
decisions based on that data
Highlight when models go wrong
(anomaly detection, bias, etc.)
15. Data Science Model Build, Run, Deploy and Manage
Automate AI lifecycle
Trust and Transparency in AI Models
Cloud Pak for Data
Advanced Analytics = Big Data + AI
Real Time Data Data Discovery Data Governance
Data Scientist & ML
Engineers
Data Management
team
Enterprise
Architect
Business
User
Data Science Use Cases
Dev-ops & Hybrid Cloud
16. Benefits
Accelerate data science projects by “as much as” 80% with
AutoAI
Improve your insight by bringing AI/ML models to Hadoop
and big data eco system
Empower citizen data scientists and analysts to gain as high as
400%+ returns on investment
Predict and optimize business outcomes using natural
language interface
AutoAI via IBM Cloud Pak for Data
Open by Design
Power of IBM
Research
Operationalize
data science
AutoAI
Hadoop
Execution
Engine
SPSS Modeler Decision
OptimizationAccelerate data
prep and model
development with
automated AI
Push AI to Cloudera
or Hortonworks
Build schedules
and plans based on
prediction
Prepare and
visualize data.
Build Models
Multiple capabilities | Consumption based | Trade-up from SPSS Available
++ +
AutoAI: Leaderboard
Base
17. User Benefits of using AutoAI
Build models faster
Automate data preparation
and model development
Jump the skills gap
No coding? No problem – get
started with a couple clicks
Discover more use
cases
Supercharge collaboration with
AI everywhere to disrupt and
transform
Find signal from noise
Auto-feature engineering makes it
easy to extract more predictive
power from your data
Rank and explore models
Quickly compare candidate pipelines
to find the best model for the job
Ready, set, deploy
Pipelines generated with AutoAI can
be deployed to REST APIs with one
click