Kogentix Automated Machine Learning Platform
Kogentix Automated Machine Learning Platform is the only solution we have seen that runs natively on Spark and includes all of the elements required to build and run a machine learning application.
Sponsored Post.
With artificial intelligence and machine learning in the mainstream media virtually every day, most organizations are aware of the importance of these emerging capabilities in establishing or maintaining a competitive advantage. Of course, the first hurdle is to identify a use case and business value where AI is applicable. Once the objective is defined, the implementation options are often not ideal.
One path is to establish a capability by hiring data engineers, data scientists, and application developers to build solutions from scratch. Given the scarcity of talent, this path can be slow, expensive, and high risk.
Another path is to access off the shelf applications that already incorporate AI, often via Software as a Service (SaaS) offerings. These solutions are almost turnkey, but rely on "black box" algorithms that offer little transparency, are limited in their ability to enable customization, and offer relatively minimal differentiation since the same solution is available to all competitors in a given domain. In the worst case, market leaders deliver a volume of data to the SaaS provider, enabling that provider to more effectively train and refine models -shifting the value of the data from the customer to the supplier.
There is a third path, exemplified by innovative startup organizations like Kogentix.
Kogentix offers an integrated platform for the development of AI applications, addressing the skills gap while enabling customers to establish an independent capability to innovate and maintain control and ownership of their data and their machine learning workflows. Maybe just as importantly, Kogentix offers not just a software platform, but also professional services to help customers kick-start their AI initiatives and ensure the initial projects yield compelling business results.
Fig. 1: Kogentix Automated Machine Learning Platform (AMP)
We recently had an opportunity to review the Kogentix Automated Machine Learning Platform (AMP) to evaluate how it addresses the challenges faced by organizations pursuing machine learning projects. Kogentix describes the purpose of AMP as enabling "not just better models, but better business applications". They appear to have delivered on an ambitious goal - Kogentix AMP is one of many emerging machine learning platforms, but it is the only solution we have seen that runs natively on Spark and includes all of the elements required to build and run a machine learning application. Kogentix AMP is built around the natural development life cycle of a machine learning application, starting with data ingestion and data discovery; moving into data wrangling and feature engineering; then supporting model development, training, and validation; and finally supporting deployment and monitoring.
Fig. 2: Kogentix AMP Dashboard.
One of the strongest advantages of Kogentix AMP is its flexibility. Data can be ingested and models can be run in both batch and streaming modes. There is a broad range of automated data discovery options, including several types of correlation, requiring no coding or scripting (which is true for virtually all features in AMP). For data wrangling and feature engineering, there are dozens of logical and mathematical operators available, with several choices for feature reduction. From a modeling perspective, AMP supports not just traditional correlation and regression, but also time-series and graph algorithms. An upcoming release will support deep learning as well.
Kogentix has a heritage of building enterprise data applications, and that experience is evident in the operational capabilities of AMP. Almost all output of AMP, from both an application development perspective and a production runtime perspective, is logged for lineage and comparison. For instance, feature relevance is tracked over time so organizations can determine which data is becoming more or less important to their desired business outcome. One of the key gaps in most data science platforms is in ongoing monitoring. AMP is designed to support a closed loop - actual results are fed back to the platform for comparison to predicted results. Where the real world diverges, alerts can be set to recalibrate models, or even to switch to a more accurate model that can run in parallel to the model of record.
Comparing various runs
Kogentix has a passion for creating actual business results out of their machine learning workflows. AMP incorporates a rules engine, enabling organizations to trigger specific actions, often in a separate system of engagement such as a sales force automation tool, a campaign management tool, or a maintenance scheduling application. Kogentix also supports a broad range of visualization solutions for reporting and dashboarding, but their primary focus is on turning insights into action.
Kogentix understands that organizations are in different stages of the journey to harness data and machine learning, and their consulting capabilities help customers move ahead no matter where they're starting from. The AMP solution itself has been guided by real world end to end solutions. They are not just a team of data scientists building tools for data scientists - they are more of an enterprise application company that leverages machine learning and AI.
The unique combination of technology and skills is probably the one reason Kogentix stands out in the quality and quantity of enterprise use cases they have delivered. A large pharma company recently used AMP to define optimal sales actions to increase the efficiency of their sales force. An international bank used AMP to better understand their retail banking customers and replaced a legacy campaign management system with a modern AI based application. They are currently working on using the same technology to improve fraud detection and avoidance. A leading trucking firm is using AMP to better predict service strategies based on error codes. These results prove that with technology available from firms like Kogentix, organizations don't need to compromise or delay in building their own roadmap of compelling AI solutions.
For more information, visit kogentix.com.
One path is to establish a capability by hiring data engineers, data scientists, and application developers to build solutions from scratch. Given the scarcity of talent, this path can be slow, expensive, and high risk.
Another path is to access off the shelf applications that already incorporate AI, often via Software as a Service (SaaS) offerings. These solutions are almost turnkey, but rely on "black box" algorithms that offer little transparency, are limited in their ability to enable customization, and offer relatively minimal differentiation since the same solution is available to all competitors in a given domain. In the worst case, market leaders deliver a volume of data to the SaaS provider, enabling that provider to more effectively train and refine models -shifting the value of the data from the customer to the supplier.
There is a third path, exemplified by innovative startup organizations like Kogentix.
Kogentix offers an integrated platform for the development of AI applications, addressing the skills gap while enabling customers to establish an independent capability to innovate and maintain control and ownership of their data and their machine learning workflows. Maybe just as importantly, Kogentix offers not just a software platform, but also professional services to help customers kick-start their AI initiatives and ensure the initial projects yield compelling business results.
Fig. 1: Kogentix Automated Machine Learning Platform (AMP)
We recently had an opportunity to review the Kogentix Automated Machine Learning Platform (AMP) to evaluate how it addresses the challenges faced by organizations pursuing machine learning projects. Kogentix describes the purpose of AMP as enabling "not just better models, but better business applications". They appear to have delivered on an ambitious goal - Kogentix AMP is one of many emerging machine learning platforms, but it is the only solution we have seen that runs natively on Spark and includes all of the elements required to build and run a machine learning application. Kogentix AMP is built around the natural development life cycle of a machine learning application, starting with data ingestion and data discovery; moving into data wrangling and feature engineering; then supporting model development, training, and validation; and finally supporting deployment and monitoring.
Fig. 2: Kogentix AMP Dashboard.
One of the strongest advantages of Kogentix AMP is its flexibility. Data can be ingested and models can be run in both batch and streaming modes. There is a broad range of automated data discovery options, including several types of correlation, requiring no coding or scripting (which is true for virtually all features in AMP). For data wrangling and feature engineering, there are dozens of logical and mathematical operators available, with several choices for feature reduction. From a modeling perspective, AMP supports not just traditional correlation and regression, but also time-series and graph algorithms. An upcoming release will support deep learning as well.
Kogentix has a heritage of building enterprise data applications, and that experience is evident in the operational capabilities of AMP. Almost all output of AMP, from both an application development perspective and a production runtime perspective, is logged for lineage and comparison. For instance, feature relevance is tracked over time so organizations can determine which data is becoming more or less important to their desired business outcome. One of the key gaps in most data science platforms is in ongoing monitoring. AMP is designed to support a closed loop - actual results are fed back to the platform for comparison to predicted results. Where the real world diverges, alerts can be set to recalibrate models, or even to switch to a more accurate model that can run in parallel to the model of record.
Comparing various runs
Kogentix has a passion for creating actual business results out of their machine learning workflows. AMP incorporates a rules engine, enabling organizations to trigger specific actions, often in a separate system of engagement such as a sales force automation tool, a campaign management tool, or a maintenance scheduling application. Kogentix also supports a broad range of visualization solutions for reporting and dashboarding, but their primary focus is on turning insights into action.
Kogentix understands that organizations are in different stages of the journey to harness data and machine learning, and their consulting capabilities help customers move ahead no matter where they're starting from. The AMP solution itself has been guided by real world end to end solutions. They are not just a team of data scientists building tools for data scientists - they are more of an enterprise application company that leverages machine learning and AI.
The unique combination of technology and skills is probably the one reason Kogentix stands out in the quality and quantity of enterprise use cases they have delivered. A large pharma company recently used AMP to define optimal sales actions to increase the efficiency of their sales force. An international bank used AMP to better understand their retail banking customers and replaced a legacy campaign management system with a modern AI based application. They are currently working on using the same technology to improve fraud detection and avoidance. A leading trucking firm is using AMP to better predict service strategies based on error codes. These results prove that with technology available from firms like Kogentix, organizations don't need to compromise or delay in building their own roadmap of compelling AI solutions.
For more information, visit kogentix.com.
