Industrial IoT

Unlocking the Power of AI & IoT with Mnubo’s AIoT Studio

4 min readAccelerate the Productization of IoT Specific AI/ML with Mnubo’s SmartObjects AIoT Studio   [...]

4 min read

Accelerate the Productization of IoT Specific AI/ML with Mnubo’s SmartObjects AIoT Studio


Artificial Intelligence (AI) and Machine Learning (ML) are powerful tools for building complex IoT systems. But launching a new IoT project and managing the full life-cycle of AI/ML presents a number of challenges, and very often they are not of the nature and scale initially planned. Mnubo’s AIoT Studio aims to overcome this challenge, empowering Data Scientists with the appropriate tools to work in a product environment.

The Truth Behind AI/ML Code

While the development of AI/ML is inherently challenging – training sets, availability of data scientists, domain expertise, etc – from a product point of view, the AI/ML code makes up a very small portion of the overall solution. Only a small fraction of real-world AI/ML systems are composed of the AI/ML code. The surrounding expertise and infrastructure is vast and complex, as depicted in the diagram below. The paper “hidden technical debt in machine learning systems“, where the below diagram is first referenced, provides an interesting angle to this topic as well. 

Diagram of hidden technical debt in machine learning systems

Figure 1.0 Only a small fraction of real-world AI/ML systems are composed of the AI/ML code

In fact, when it comes to commercial solutions, deploying and running AI/ML code against live streaming data as well as maintaining it over time, requires a specific level of expertise and goes well beyond the original challenges of AI/ML. Add the need for live collaboration between experts, corporate-wide centralization and circulation of this intellectual property (IP), and you have yourself a very expensive puzzle to solve.

About Mnubo AIoT Studio

It is with this complexity in mind that Mnubo developed and released its SmartObjects AIoT Studio. This fully-managed IoT-focused AI/ML studio allows customers to develop, train/maintain, deploy, and run AI/ML models as well as ad-hoc algorithms against live streaming data. Using microservices and a function-as-a-service architecture, Mnubo’s AIoT Studio unscrambles and abstracts the complexity of building and maintaining the infrastructure required to power production ready AI/ML at scale.

Mnubo AIoT Studio: An Overview

  • Develop – access a Python AIoT studio while leveraging, when needed, additional open source libraries like scikit-learn, Keras, Spark MLlib, FB Prophet, etc.; add your own dependency set(s) if required and develop your custom intellectual property.
  • Train/Maintain – query any archived and/or live streaming IoT data to train, test and maintain your code and models.
  • Deploy – productize and make your code available to any internal/external processes and applications.
  • Run – access and execute any deployed code via a simple function-as-service JSON REST call.

Features and Benefits

No ETL data extraction

  • Query any archived and/or live streaming IoT data stored in SmartObjects using an extensive JSON query language

Access any AI/ML framework or libraries

  • Direct integration with SmartObjects’ native IoT libraries (e.g. anomaly detection, time to target, session analytics, timeseries prediction, scoring, etc.)
  • Easy access to external libraries (public and private) through dependency sets (e.g. scikit-learn, Keras, Spark MLib, FB Prophet, etc.)

Centralize your IP and share worldwide

  • Version your code and share with your colleagues worldwide
  • Centralize all your corporate IoT-specific IP in one repository and reuse/ distribute as often as required
  • Bring your own model if your development was already done using a third-party tool
  • Store binaries separately from your code when required

One click deployment process

  • Access a sandbox environment and test-before-you-deploy feature
  • Train, test and refine against archived and/or live streaming data
  • One click productizing and deployment process via function-as-a-service
  • Monitored and unified logging system

Scale as you grow architecture

  • Assign the right number of CPUs and amount of RAM when deploying to balance speed versus costs
  • Schedule the execution of your code to store results in separate indexes for fast retrieval by Web and Mobile Apps

Screenshot of Mnubo's AIoT Studio

Figure 2.0: AIoT Studio – Example of Temperature Prediction using TensorFlow

Business Use Cases

Many of Mnubo’s IoT customers are already leveraging the power of SmartObjects AIoT Studio to:

Schedule Proactive Maintenance
Use artificial intelligence models to predict machine downtime and schedule proactive maintenance. Identify impending faults early to proactively counteract product malfunctions and reduce costly emergency repairs

Screenshot of Mnubo's SmartObjects platform

Figure 3.0 Identify impending faults early to proactively counteract product malfunctions

Enable Preventative Replenishment
Use advanced data science and machine learning models to enable targeted preventative actions. Predict time to action for equipment requiring replenishment/replacement. Help field technicians with inventory and resolution planning.

Focus R&D Spend
Identify anomalies before products are being deployed and updated. Determine which equipment has the most faults and alerts, what the trend is over time, how does this vary across product versions.

Screenshot of Mnubo's Anomaly Detection dashboard

Figure 4.0 Identify anomalies before products are being deployed and updated.

Improve Service Efficiency
Visualize critical sensor readings and asset states in real-time to reduce downtime, service calls and truck rolls. Create new service offerings, empower dealer networks with condition-based monitoring and automated reports to improve service efficiency

Increase Engagement & Reduce Churn
Understand how customers are interacting with your connected product. Group customers based on their usage behaviour, and profile them to facilitate a faster adoption rate. Track the entire connected product life-cycle to reinforce brand engagement.

Mnubo's SmartObjects Insights

Figure 5.0 Understand how customers are interacting with your connected product

Learn more about SmartObjects today.