Deep Link Analytics Empowered by AI + Graph + Verticals
0 likes85 views
The document discusses the role of AI in various industries and highlights the importance of deep link analytics for integrating and optimizing data across multiple verticals. It emphasizes that many AI projects fail to go live and provides insights into how businesses can leverage AI to enhance customer understanding and operational efficiency. Landing AI's vision platform aims to empower organizations to create, deploy, and scale AI solutions effectively, focusing on customer 360 and real-time data analytics.
Deep Link Analytics Empowered by AI + Graph + Verticals
- 1. Deep Link Analytics Empowered by AI + Graph + Verticals Dongyan Wang VP AI Transformation, Landing AI
- 2. ANDREW NG CEO and Founder • Co-Founder of Coursera • Founding lead of Google Brain • Former Baidu Chief Scientist • Stanford Computer Science Adjunct Professor
- 3. © 2020 LANDING AI | 3 Retail Travel Transport & Logistics Automotive & Assembly Basic Materials Advanced Electronics/Semiconductors Healthcare Systems and Services High Tech Telecom Oil & Gas Agriculture $0.8T $480B $475B $405B $300B $291B $267B $267B $174B $173B $164B Source: https://www.mckinsey.com/featured-insights/artificial-intelligence/visualizing-the- uses-and-potential-impact-of-ai-and-other-analytics Impact by Economic Sector AI 13 TUSD AI value creation by 2030 $13TRILLION
- 4. © 2020 LANDING AI | 4 Landing AI Focus Vision Platform • Manufacturing Visual Inspection • eCommerce • Healthcare… Deep Link Analytics • Customer 360 • Recommenders • Telecom, E-Commerce, retail… • Platform (Upcoming) AI Transformation Partnership Projects, Strategy, Hiring, Training, Communication
- 5. 75% of AI projects do not go live PoC /Experiment Works on researcher’s laptop. May be publishable. Deployment Running “live” in production and creating value for at least 1 user. Scale up Creating significant business value. HARD HARD HARD
- 6. © 2020 LANDING AI | 6 ML Code Customers Need More Than ML Code Defect Definition & Ambiguity Resolution Detect Image Acquisition Issues Early On Efficient, Batch Data Labeling ML Code Data Verification Defect, Data Analytics Data, Model Version Control Process Management Tools Detect Model Performance Degradation Environment Change Detection Create data for Rare Defects Only 5% - 10% of AI projects are ML code Landing AI is good at building a customized stack to serve different verticals
- 7. © 2020 LANDING AI | 7 Landing AI Vision Platform: Visual Inspection, Healthcare… Enable customers to create, deploy, scale AI solutions with a close-loop platform
- 8. © 2020 LANDING AI | 8 Manufacturing is the Perfect Playground for AI Applications
- 9. © 2020 LANDING AI | 9 • Faster: unify data across silos & generate business value • Deeper: understand complex data relationship • Bigger: navigate huge amount, ever growing data • Optimized for vertical business: analytics for my business problem! Customers Needs – Deep Link Analytics even after years of work in BI, EDW, Analytics…
- 10. © 2020 LANDING AI | 10 • How can I sell more to existing customers? • How can I identify new customers? • What do my customers need me for today and future? • Who are my customers' suppliers or what brands do they like? • How I can retain my customers and prevent churn? • How do I price my products? • What products or services will my customers buy? • How efficiently can I deliver products to my customers? • How efficiently can I forecast what products are needed and when? • others ( i.e., Unified ID, customer 360 can expand to healthcare, transportation and many) Generate Business Value Fast & Profound!
- 11. © 2020 LANDING AI | 11 The Power of AI + Graph + Vertical Data/Analytics Graph Database Architecture Specialized Vertical AI Algorithms Customer, Sales, Product 360 Deep Data Connection Data Injection Actionable AI Analytics Optimized Vertical AI AnalyticsData Injection Landing AI Deep AI Analytics Specialized Vertical AI Algorithms ID Unification, fraud Detection Specialized Vertical AI Algorithms e-Comm / Retail / Telecom Recommendations High performance querying Pre-optimized schema for vertical & AI analytics Quickly connect siloed data Product Specialization Knowledge Networking
- 12. © 2020 LANDING AI | 12 • Integrate high volume data at high speed and analyze in real-time • Organize data to reveal deep relationship for decisions • Pre-build schemas using decades of data & business expertise • Application ML - algorithms for targeted business solutions • Ongoing Model Management – versioning, monitoring, CI/CD • Domain expertise, architecture, customizations What is Deep Link Analytics?
- 13. © 2020 LANDING AI | 13 Landing Deep Link Analytics: AI + Graph + Vertical Engine Client Data Graph Vertex per record ingest high volume at high speed from enterprise systems Landing Analytics SubGraph client-agnostic application features (derived queries) for AI Modeling Landing Data-Model Subgraph algorithms to develop deep data fabric using pre-defined Landing’s schema Unify data silos + real-time + high performance ID Unification Process Customer & Product 360 X & Up-Sell Recommender Demand Sense, Prediction MLDeploy+CICDAutomation Vertical App Centric ML models
- 14. © 2020 LANDING AI | 14 Deep Link Analytics: AI + Graph + Verticals Quickly unify siloed data with deep data relationships, integrated across multiple data sources Organize data using predefined vertical specific schemas for feature engineering & analytics Integrated Multi Model framework using vertical specific ML/DL models & graph algorithms Customizable APIs to integrate AI analytics with customer apps Automation framework with CI/CD capabilities Model Deployment Data Pipelines Data Organization Multi-model Algorithms Custom APIs AI Powered Deep Analytics
- 15. © 2020 LANDING AI | 15 Core Business Value • Integrate & organize siloed customer data to deep knowledge network • Knowledge network & sharing, data connections, app intelligence • in an efficient way to drive critical decision making • High performance query • Understand deeper relations for focused verticals, starting from customer 360 • Across consumers, products, brands & demand • Knowing consumers better - consumers emotions and sentiments • Predicting demand and deliver services for improved customer satisfaction • Transform data to actions using scientific algorithms • Customer 360 / Recommendations for x-sell & up-selling products & services • Demand Sensing, Forecasting / Supply Chain
- 16. © 2020 LANDING AI | 16 Deep Link Analytics Use Cases Customer 360 Customer churn prevention Lifetime value prediction Recommendation engines Customer sentiment analysis Customer Journey Predictive analytics Fraud detection Network management Route optimization Price optimization ID Unification: Use deep link analytics to unite/dedupe different users ID across data silos, and differentiate same or different users Customer 360: Understand customers full persona across various engagement and interactions captured as part of their engagement with a product, service or support for sentiments/interests X-sell / up-sell: know customer better for their needs to buy the right product at the right time Route Optimization: Vehicle history for routes, repairs, fuel rate, usage, segments (routes), schedule activities (scheduled vs. actual). And customer 360 can be integrated to add more value in terms of optimizing based on deeper understanding of customer travel patterns across time, place and events of interest and family structure. Predictive Maintenance: Based on vehicles history in terms of various factors like repairs, breakdowns, services, total cost of ownership including consumption, usage etc., Predictive Analytics: Ability to understand historical failures pattern and learn to predict future failures based on specific identifiable signals over time Fraud Detection: Ability to understand and identify potential fraud/ risk associated with transactions based on deep data relationships and risk pattern identifiable within the data network
- 17. 10-100s Products 10-100s Defects/ Data Patterns x 100-1000s Labeled Data x Gain Control of Your AI Have to Manage Compounding Complexity of AI Apps?
- 18. Offline PoC Deploying AI Needs Model Performance Management Gain Control of Your AI Time Performance Online PoC AI Rampup Track and Adjust 100% Track Performance Drops Detect Data Drifts Continous Learning Continous Deployments HLP Human in the Loop (Virtuous Cycle) EasyCases Hard Cases
- 19. Find the right use cases Be ready for data Plan for deployment not PoC Gain control of your AI AI Around the Corner.
- 20. © 2020 LANDING AI | 20 Thank you. Dongyan Wang Dongyan@landing.ai 1-650-695-1878