Open Forum

AIoT doesn't have to be complicated.

At yesterday's Montreal Elastic Meetup, I presented how we achieved meaningful and relevant anomaly detection of ambient illuminance data, with neither expert knowledge about IoT or AI:



------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive
------------------------------

Very well said, Jeffrey... especially with the abundance of commidity sensors, SoC computers, and open source libraries (both embedded and web-apps), implementing a solution has become quite easy. I think the challenge is to carefully define the problems... in this world, if the problem is properly defined, it is like already 75% solved!

------------------------------
Suvojit Ghosh
CEO
FYELABS
Hamilton ON
------------------------------

Original Message:
Sent: 06-15-2022 13:12
From: Jeffrey Dungen
Subject: An 'illuminating' example of AIoT

AIoT doesn't have to be complicated.

At yesterday's Montreal Elastic Meetup, I presented how we achieved meaningful and relevant anomaly detection of ambient illuminance data, with neither expert knowledge about IoT or AI:

Pareto Anywhere open source middleware running on a Raspberry Pi discovered and interpreted the ambient illuminance data from a Bluetooth sensor that happened to be nearby, and forwarded this data to an Elasticsearch instance running in the Elastic Cloud.  The only manual intervention was instructing the Elastic Machine Learning module to consider illuminance data for anomaly detection, and the result is what you see above!

This is a perfect example of an AIoT quick win that can be replicated with ease at marginal cost, and applied to just about any timeseries sensor data.  It's just a matter of selecting and assembling off-the-shelf elements as shown in the arrow above.  Hope you find that as 'illuminating' as I do!

------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive
------------------------------

Thanks for the follow-up @Suvojit Ghosh, you are right that a well-defined problem is very much the difference between a technology demo and a solution.  What I'll add is that the choice of off-the-shelf and open source technologies to create a solution from a well-defined problem is not always obvious, nor is selecting and observing the standards and best practices that will contribute to a long-term solution in terms of operation and maintenance.  There's still a lot of room for improvement in our industry to encourage sustainable and interoperable solutions.​

------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive
------------------------------

Original Message:
Sent: 06-18-2022 22:55
From: Suvojit Ghosh
Subject: An 'illuminating' example of AIoT

Very well said, Jeffrey... especially with the abundance of commidity sensors, SoC computers, and open source libraries (both embedded and web-apps), implementing a solution has become quite easy. I think the challenge is to carefully define the problems... in this world, if the problem is properly defined, it is like already 75% solved!

------------------------------
Suvojit Ghosh
CEO
FYELABS
Hamilton ON

Original Message:
Sent: 06-15-2022 13:12
From: Jeffrey Dungen
Subject: An 'illuminating' example of AIoT

AIoT doesn't have to be complicated.

At yesterday's Montreal Elastic Meetup, I presented how we achieved meaningful and relevant anomaly detection of ambient illuminance data, with neither expert knowledge about IoT or AI:

Pareto Anywhere open source middleware running on a Raspberry Pi discovered and interpreted the ambient illuminance data from a Bluetooth sensor that happened to be nearby, and forwarded this data to an Elasticsearch instance running in the Elastic Cloud.  The only manual intervention was instructing the Elastic Machine Learning module to consider illuminance data for anomaly detection, and the result is what you see above!

This is a perfect example of an AIoT quick win that can be replicated with ease at marginal cost, and applied to just about any timeseries sensor data.  It's just a matter of selecting and assembling off-the-shelf elements as shown in the arrow above.  Hope you find that as 'illuminating' as I do!

------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive
------------------------------

Original Message:
Sent: 6/20/2022 11:43:00 AM
From: Jeffrey Dungen
Subject: RE: An 'illuminating' example of AIoT

Thanks for the follow-up @Suvojit Ghosh, you are right that a well-defined problem is very much the difference between a technology demo and a solution.  What I'll add is that the choice of off-the-shelf and open source technologies to create a solution from a well-defined problem is not always obvious, nor is selecting and observing the standards and best practices that will contribute to a long-term solution in terms of operation and maintenance.  There's still a lot of room for improvement in our industry to encourage sustainable and interoperable solutions.​

------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive
------------------------------

Original Message:
Sent: 06-18-2022 22:55
From: Suvojit Ghosh
Subject: An 'illuminating' example of AIoT

Very well said, Jeffrey... especially with the abundance of commidity sensors, SoC computers, and open source libraries (both embedded and web-apps), implementing a solution has become quite easy. I think the challenge is to carefully define the problems... in this world, if the problem is properly defined, it is like already 75% solved!

------------------------------
Suvojit Ghosh
CEO
FYELABS
Hamilton ON

Original Message:
Sent: 06-15-2022 13:12
From: Jeffrey Dungen
Subject: An 'illuminating' example of AIoT

AIoT doesn't have to be complicated.

At yesterday's Montreal Elastic Meetup, I presented how we achieved meaningful and relevant anomaly detection of ambient illuminance data, with neither expert knowledge about IoT or AI:

Pareto Anywhere open source middleware running on a Raspberry Pi discovered and interpreted the ambient illuminance data from a Bluetooth sensor that happened to be nearby, and forwarded this data to an Elasticsearch instance running in the Elastic Cloud.  The only manual intervention was instructing the Elastic Machine Learning module to consider illuminance data for anomaly detection, and the result is what you see above!

This is a perfect example of an AIoT quick win that can be replicated with ease at marginal cost, and applied to just about any timeseries sensor data.  It's just a matter of selecting and assembling off-the-shelf elements as shown in the arrow above.  Hope you find that as 'illuminating' as I do!

------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive
------------------------------

Very well said, both of you @Jeffrey Dungen and @Suvojit Ghosh :

- Defining the problem​​ is getting at least half-way to fining the right solution.

I'd encourage enterprises to team up with some trusted experts. As somebody at BOSH once said "you cannot do IoT ( and by extension AIoT ) alone"

------------------------------
Salvatore Cimmino
------------------------------

Original Message:
Sent: 06-20-2022 11:42
From: Jeffrey Dungen
Subject: An 'illuminating' example of AIoT

Thanks for the follow-up @Suvojit Ghosh, you are right that a well-defined problem is very much the difference between a technology demo and a solution.  What I'll add is that the choice of off-the-shelf and open source technologies to create a solution from a well-defined problem is not always obvious, nor is selecting and observing the standards and best practices that will contribute to a long-term solution in terms of operation and maintenance.  There's still a lot of room for improvement in our industry to encourage sustainable and interoperable solutions.​

------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive

Original Message:
Sent: 06-18-2022 22:55
From: Suvojit Ghosh
Subject: An 'illuminating' example of AIoT

Very well said, Jeffrey... especially with the abundance of commidity sensors, SoC computers, and open source libraries (both embedded and web-apps), implementing a solution has become quite easy. I think the challenge is to carefully define the problems... in this world, if the problem is properly defined, it is like already 75% solved!

------------------------------
Suvojit Ghosh
CEO
FYELABS
Hamilton ON

Original Message:
Sent: 06-15-2022 13:12
From: Jeffrey Dungen
Subject: An 'illuminating' example of AIoT

AIoT doesn't have to be complicated.

At yesterday's Montreal Elastic Meetup, I presented how we achieved meaningful and relevant anomaly detection of ambient illuminance data, with neither expert knowledge about IoT or AI:

Pareto Anywhere open source middleware running on a Raspberry Pi discovered and interpreted the ambient illuminance data from a Bluetooth sensor that happened to be nearby, and forwarded this data to an Elasticsearch instance running in the Elastic Cloud.  The only manual intervention was instructing the Elastic Machine Learning module to consider illuminance data for anomaly detection, and the result is what you see above!

This is a perfect example of an AIoT quick win that can be replicated with ease at marginal cost, and applied to just about any timeseries sensor data.  It's just a matter of selecting and assembling off-the-shelf elements as shown in the arrow above.  Hope you find that as 'illuminating' as I do!

------------------------------
Jeffrey Dungen
Co-founder & CEO of reelyActive
------------------------------
​