Slides from my keynote at the Santa Clara executive event on Oct 26, 2017 attended by senior delegates from large Semiconductors, consumer electronics & Network firms. I shared my perspectives on fundament building blocks of Artificial Intelligence and use cases for enterprises.
Linked Data in Production: Moving Beyond Ontologies
Artificial Intelligence - the next generation of automation
1. Bimal Tripathi Oct 2017
Bimal Tripathi
Sr Director
Brocade Communications
Trends and Opportunities
In Information Technology
2. Bimal Tripathi Oct 2017
Enterprises Artificial Intelligence
TOPICS
• Enterprise adoption
• Demystifying AI
• Business Case for Enterprise AI Strategy
• Enterprise AI Roadmap
• Ethics
3. Bimal Tripathi Oct 2017
Enterprise adoption…
Enterprises do not deal with smart
cars, natural language processing,
Computer vision and photo tagging
but…
… would benefit from AI by using
cumulative enterprise digital
experience to
• Grow revenue
• Improve products
• Enhance Customer Engagement
• Optimize operating efficiencies
4. Bimal Tripathi Oct 2017
2
Demystifying AI - How it works?
Computer learns from data
1
3
Recognizes patterns in the historic data
higher dimensions and larger data volumes
Builds models (statistical rules ) to predict
Applied the rules on new data to make
decisions
5. Bimal Tripathi Oct 2017
Historic Data
Pattern
Rule
48.5 + 0.05 * Revenue
Example 1: Supervised Learning – Regression
What is the right deal size for a customer given Revenue $6,040 M?
6. Bimal Tripathi Oct 2017
Will App crash?
CPU Utilization : 70%
Memory Utilization : 85%
Historic Data
Pattern
Rule – A decision tree
Example 2: Supervised Learning – Decision Tree
Predict Service disruption - is the service availability at risk?
7. Bimal Tripathi Oct 2017
Historic Data
Rule – a neural network
Example 3: Supervised Deep Learning (Neural Networks)
Lead quality prediction – what leads should internal sales reps work on?
8. Bimal Tripathi Oct 2017
Personalized offers
and messaging
Example 4: Unsupervised Learning
Customer Segmentation – group “look alike customer” for the next campaign
CUSTOMER Data CLUSTERED DATA
9. Bimal Tripathi Oct 2017
Web Traffic
Response time
Example 5: Unsupervised Learning
Anomaly detection – Real or False Alarm?
10. Bimal Tripathi Oct 2017
Business case for Enterprise AI Strategy
Last 3 decades of business innovation…
ERP -> Internet -> Analytics -> Virtualization -> Cloud -> Mobile
Is AI the next frontier?
Strategic Goals
Grow revenue
Enhance Customer Engagement
Improve products
Optimize operating efficiencies
Execution
Success Metrics
Process Owners
Governance
Infrastructure
Data Science Lab
Data Architecture
At scale Computing
HDFS/SPARK/PYTHON
11. Bimal Tripathi Oct 2017
Sales/CRM Product
Operations
Marketing &
eCommerce
Information
Technology
People
Operations
Enterprise Artificial Intelligence
Use case for Enterprise functions
Bot and
assistants
Opportunity
ranking
Pricing
Customer churn
Support
readiness
Sentiment
Product Failure
Inventory Mgmt
Forecasting
Anomaly
detection
Lead scoring
Recommendations
Promotions
Segmentation
Dynamic pricing
Buyer intent
Cyber Security
Help Desk
automation
Data Quality
System
availability
Recruiting
Engagement
Attrition
Data sources:
• In-house applications – ERP, Web, CRM, products
• Data Warehouse/Data Lake/Big Data Clusters
• 3rd party data providers
13. Bimal Tripathi Oct 2017
PREPARED FOR AN ENTERPRISE AI STRATEGY?
1- Align and
Plan
2- POC &
Simulation
3- Design &
Build
4- Track &
Monitor
Establish ML goal
Build domain knowledge
and engage SME
Build hypotheses and
experiment
Determine the sources
of data
Define success criteria and
performance metrics
Find strongest
predictors of event
Select model and run
train/test validation
Collect and enrich data
for model building
Iterate and pick best
model
Design deployment
architecture for scale
Select Machine
Learning Technologies
Identify data sources and
needed process changes
Deploy new data taps
Institute systemic data
Quality validations
Deploy for production
usage
Monitor Prediction accuracy
& adjust Model parameters
Study Performance
Trends
Root cause and correct
performance detractors
Assess cycle time to
results throughput SLAs
Conduct capacity review
and technology upgrades
Sample tools and technology:
Dashboards
Visualization Tools
Monitoring tools
Sample tools and technology:
Database, Excel
R, Python, SAS, RapidMiner,
TensorFlow, H2O, Spark Mlib,
Amazon AML
Sample tools and technology:
Dashboards
Visualization tools
Roadmap to implementing Enterprise Machine Learning
Sample tools and technology:
Data repository (RDBMS, HDFS)
Python, Java, Spark, R
Infrastructure: AWS, Google,
SFDC, Oracle, SAP, IBM, MSFT
14. Bimal Tripathi Oct 2017
Ethics of Artificial Intelligence
❖Bots versus humans
❖Social Economics
❖Employee privacy
❖Customer profiling