MLSEV Virtual. One Platform to Rule Them All
0 likes226 views
A1 Digital focuses on providing an integrated data platform for enterprises, specializing in IoT, machine learning, cloud, and security solutions. The platform supports advanced analytics and seamless data integration for various industries, with features like no-code interfaces, personalized reporting, and automated deployments. Notable use cases include predictive maintenance in energy, churn prediction in retail, and fraud detection in housing subsidies.
MLSEV Virtual. One Platform to Rule Them All
- 1. One platform to rule them all 26.03.2020, Francis Cepero
- 2. 2 Who we are A1 Digital – IoT, ML, Cloud and Security Focus 3 Headquarters Vienna, Munich and Lausanne - present in 10 countries 180 Employees More than 500 international customer projects Affiliated Group A1 Telekom Austria Group & America Móvil Your Partner for Cloud, ML, IoT and Security
- 3. 3 Implement cloud or on-premise ML Advanced Analytics Powered by Enterprise Scale Reporting & BI Dashboarding Large Scale Data Processing & Analysis Seamless Integration Images Social Media IoT Data Text Data Share Actionable Insights Predict Business Outcomes Marketing Management Customers Empower Business Ingest from diverse formats and sources The Goal*: One integrated Data Platform for the enterprise *validated in industry conversations
- 4. 4 Machine Learning • Topic modeling: Extraction of concepts, entities and categories • Predictions, Classifications, Clustering, Anomaly Detection, Associations, etc • Feature Engineering • Time Series Analysis • Automatic selection of algorithms/models Data Discovery • Data ingestion of all major data sources • Intuitive, no code/drag-and-drop interface • OOTB data quality summarization data processing and advanced analytics techniques available Power to the Business • Business Applications can be built on top • High granularity wrappers and scripted flows • Programmable and traceble flows and models • Artifacts are exportable and integratable at data, algorithm, process and model level BI & Reporting • Personalized, interactive data visualizations and dashboards • Design tool for developers and end users to build reports and dashboards • High performance, scalable Infrastructure • Integration and APIs first • Zero install: Autoscalable**, autodeployable • Fully-automated, Abstracted Serverless • Cloud, Edge, OnPremise deployments Security • Single sign-on • Secure data • Anonymizable/Pseudonymizable • Secure communications One Platform: complex requirements, very difficult to build* *validated in industry projects **Sauron says hello
- 5. 5 One platform: Many different industries
- 6. 6 LoB Topic Construction Waste Energy OEM Retail Public Sector Facility Logistics Service Mgmt Banking HR Workload prediction x x x Resource planning x x x x x Supply chain management Product/Catalogue classification x x x x Marketing Promotion planning x Controlling Revenue forecast x x x x x x x x x x Security Fraud Detection x x Operations Demand forecast x x x x Sales Sales monitoring x x x Production Quality assurance x R&D Trainings & platform x x x x x x A1Digital delivering projects in different industries and LoBs
- 7. 7 A1Digital: Many different use cases Marketing campaign planning Cannibalization effects in retail Churn predictionHousing subsidy fraud detection Predictive maintenance Flat wheel detection Bee health monitoring Product classification in procurement Call center resource planning ML for energy tradingSales monitoring Quality insurance
- 8. Deep Dive Examples Cannibalization effects in retail Flat wheel detectionML for energy trading
- 9. 9 Industry:Energy - LoB: Trading in energy balancing markets* *tertiary markets balancing energy pool for surplus demand forecasts
- 10. 10 Industry: Retail - LoB: Promotion Planning
- 11. 11 IoT (energy autonomous) • Data collection: IoT Devices can send regularly different data from various Sensors ✓ • Flat-spots Detection: vibration is tested to successfully detect flat spots ✓ ML for Rail Industry:Rail Transportation - LoB: Asset Operations/Maintenance Data Analytics (Cloud-based) • Data Transmission: Volume of >1,5GB per Device/Day ✗ • Energy: transmission is energy consuming ✗ • Reliability: limited cellular coverage restricts data processing ✗
- 12. 12 The Solution: TinyML and BigML Embedded machine learning core On an accelerometer to run decision trees Apply BigML Using BigML capabilities for data insights, feature and model selection Find a balanced model Evaluate, compare and optimized model for deployment
- 13. 13 The Solution: Flat Spots detection with TinyML IoT Platform ML Model runs after movement for 20 seconds to detect flat-spot: • yes / no • probability Model Deployment over the air (OTA) Flat Spot warning Process test Data from real Devices Create BigML Model Develop TinyML Model Alarm Message! • Issue: Flat spot • Wagon: Nr. 12405 • Probability: 93% • Count of Alarms: 12 TinyML ▪ ‚Down-sized‘ Machine Learning Models run on microcontrollers with very low power consumption ▪ Data remains on the IoT devices and is processed & analyzed directly on the device ▪ Only single events (Flat-spot warnings) are transmitted with probability values ▪ Energy consumption remains minimal ▪ Limited cellular coverage does not affect the use case ML for Rail
- 14. 14 TinyML Journey for the customer • Use case definition • Business Impact estimation • Sensor selection Scoping • IoT devices • Microcontroller • Connectivity IoT & Hardware • Field experiment • Data collection • Data labeling • Quick analytics Experiment • Feature selection • Model optimization • Model sizing ML Development • Test deployment • Validate collected data • Simulate critical situations Validate • over the air • Evaluate & update • Connect to A1 Digital IoT Platform Deploy 2-3 weeks 4-8 weeks 4-8 weeks
- 15. Thank you