The industry-leading cloud & SDK product for collecting and analysing automotive data sets.
The key will be combination of proactivity, personality and ability to manage tasks, based upon learning about your personal data.
CloudMade’s framework allows you to maximize data value using 3 learning approaches.
A unique architecture
Our flexible frameworks architecture allows you to deploy CloudMade components within your own solution, leveraging your vehicle or smartphone sensor network and then computing onboard and/or in cloud to deliver predictions back to the device.
Adaptive Framework components
CloudMade’s Adaptive Framework allows you to build intelligent
mobility solutions faster and with more flexibility.
Select the jobs you are focusing on to find out more about the different framework
components and how they can work for you.
Feature preparation components
Toolchains that check the quality and integrity of feature data to create reliable datasets.
Processes data feeds across devices. Provides enrichment from third party sources and builds journeys for IE use.
A process to filter and report on the quality and integrity of data feeds.
Data validation reporting
Reports and alerts on data validation.
ML delivery & distribution components
Manage the creation, updates and distribution of personal profiles that enable predictions across cloud, phones and vehicles.
Inference engine scheduler
Uses standard scheduling software as a base to schedule IE jobs.
Model repository manager
A repository to store personal inferences.
A user-device Edge service for management and profile distribution.
Cloud predictions REST APIs
APIs that provide a wrapper for the execution of prediction plug-ins in the cloud for use in portals, web apps or via web-api from other devices.
A process that monitors context signals to determine whether predictions need to be updated, and if so executes them. Predictions are published and available at any time for standard context. Predictions for specific context (including what-if ad hoc requests) are available on demand via RPC like mechanisms.
An in-vehicle process that manages the local cache of user profiles from cloud or on-board learning. On-board and off-board machine learning for a user is merged with specific on-board or off-board algorithms.
User authentication and confirmation; GDPR compliancy; onboard and offboard resiliency.
User authentication service.
Data injection components
Acceptance and pre-processing of data from the automotive eco-system; passing features to events.
Data Import for appliances
A tool to load data into an appliance event feeder.
A service that accepts and processes data from multiple sources including vehicles, vehcile telematic systems, data lakes and mobile devices.
Streaming event feeder
A pre-processing stage to accept streaming data and execute feature extraction with links back to the source stream (source / index to allow inspection), and passing features to the event feeder as events.
Munic.io data processing toolchain
A toolchain to process and collect data from munic.io dongles.
A process that accepts events, manages them in a queue, implements storage managment rules and submits the stored events for synchronization with cloud systems according to business rules.
Visualization dashboards components
Web-based dashboards for production maintanance and data science research.
A dashboard for exploring profiles and data.
Validation framework dashboard
A dashboard for production maintance or data science research related to the validation framework.
Not customer facing, but provides a basis for enhancements to the driver dashboard.
Inference engines components
Use-case centric machine-learning algorithms for intelligent mobility.
Python inference API plugins
A set of plugins that provides learning (IE) and predictions for a specific domain.
Predictive routes and destinations inference engine
A job for creating predictive routes and destination profiles for vehicle predictions.
Prediction plugins 1 per engine
A portable plug-in that provides predictions for specific IEs.
Java IEs in cloud 1 per inference
A job for creating and training a ML model for specific IEs.
Machine learning API components
Components that deliver APIs for external service consumption.
Python inference API
A component that provides python API for model development in python.
Data API for prototyping (integration with sagemaker, etc)
A set of cloud-specific views that allows for working directly with journey data from within cloud-provider ML tools. Not implemented yet, but planned.
On-board learning manager
A vehicle process that manages and executes on-board inference engines to locally create inferences for inclusion in device profile and execution.
ML utility libraries
A broad set of Machine Learning algorythims tuned for use in learning and prediction on vehicle and mobile device architectures.
Geo data management components
Toolchains that enable hybrid geospatial data management processing for caching, searching and tiling services.
Map data import toolchain
Processes that convert, import or update third-party map data into a hybrid format.
Fleet learning server
An integrated set of processes that are a base for crowdsource data from various sources that feeds into hybrid map datasets (e.g. ACC usage).
Service for efficient layered geo-data distribution.
Hybrid onboard server
An on-device service that manages local cache and services clients across the device/vehicle.
A library that provides access to local hybrid cache and optionally syncs with a hybrid server, as required.
Hybrid place naming
A service that matches geo coordinates to place names using a variety of approaches. This feature can potentially integrate with proprietary services.
Hybrid streaming data management
A pre-processing stage to accept streaming data and execute feature extraction with links back to the source stream (source / index to allow inspection), and passing features to fleet learning or other hybrid storage systems.
Validation methods components
Quality checks for driver profiles and machine-learning algorithm predictions.
A set of jobs for checking the quality of machine learning algorithms.
Profile quality measurement job
A job that evaluates a driver’s profile quality using validation framework results. Enables profile suppression. Might be a part of profile builder.
CloudMade’s framework products are the result of years of development and are available right now.
If you are thinking of developing your own machine-learning framework, talk to us first before commiting valuable budget resources for internal development. Please get in touch for more information about our any of our framework components.
CloudMade’s Adaptive Framework allows you to design your intelligent vehicle architecture to maximise reusable components, then build and deploy services on top of that at your own pace.
We are constantly analysing and validating the quality of our algorithms and improving performance. If you are thinking of developing your own machine-learning inference engines or the services running on top of them, talk to us before using your own valuable in-house development resources.