Early Years, I was hopeless on the sports field and didn't excel academically in my early school years. When I was 13 years old I got bored of playing video games so I designed and wrote one.
Graduated 1991. University of the Witwatersrand BSc Engineering (Industrial) Cum Laude
Exel Engineering (1992-1993) - Consultant. My first consulting engagement got me immersed in reliability and stats. I designed and built a system that collected and analysed aircraft failure data using statistical methods.
Designed and built an application to track incoming aircraft spare parts for a new fleet of aircraft. The system recommended a pilot training schedule based on the availability of spare parts.
Ernst & Young (6mo 1994) - Contractor. Trained consultants in the use of E&Y's Information Engineering methodology to model strategy and design computer systems.
Northern Medical Aid (1994-1997) - Contractor. Designed the database for a new client/server medical administration system. The system was built using PowerBuilder and Sybase. Became team lead for the development of the claims processing module. We managed to automate most of the claims assessment process. This was my first exposure to complex rules-based systems for decision making.
Built a data warehouse along with OLAP, business intelligence and custom risk management apps. Hired and managed the data warehouse team.
Business Objects (3mo 1998) - Contractor. Did a short stint doing Business Objects implementations for Mike Bergen & Associates
Cognos (1998 - 2008) - Director of Architecture. The first products that I worked on for Cognos were prebuilt Sales and General Ledger data warehouses for SAP.
Cognos Metrics Studio. I was looking for a way to bring data warehouse data to life in a simple, engaging manner...without the need to pre-build reports. Cognos Metrics Studio was the first product that I got to conceive and architect from the ground up. Myself, our VP and the design team obsessed about every little detail of this product....and it paid off:
"Cognos metrics studio is a simple product - much fewer features than the more mature Balanced Scorecard products. Despite this, it is the product that gets rolled out more broadly than the rest." Frank Buytendijk, Gartner. This was to become a recurring theme in my work - simplifying products to make them easy to live with, easy to roll out, and consumable by lots of users.
The Applications group grew rapidly. As chief architect I oversaw the build out of a portfolio of data warehouse and BI applications for all of the major ERP systems covering additional areas: inventory, procurement, workforce, AR and AP. With this rapid growth, we developed good data warehouse and BI IP in the from of high value, repeatable business insights. There were also significant growth pains:
I led a re-architecture effort. The new metadata driven approach shaved months off the development time for each app, made apps easy for customers to customize and allowed us to upgrade customized apps. As an added benefit, the new architecture connected the apps for cross functional reporting- providing seamless access to any data across the portfolio.
IBM Acquires Cognos > Architect, Senior Technical Staff Member. As part of a new initiative to drive business growth through AI, I was asked to architect a suite of AI solutions for industrial, energy and sales use cases. The first release of the product was collection of SPSS assets bundled into a new management interface. After proving the concept, I was given access to a team of data scientists and we set about improving the predictive models and fully "operationalizing" them behind IBM Message Broker.
Predictive Quality: A modern version of the Taguchi Methods
Cognitive Plant Advisor. In collaboration with a team of IBM researchers, we productized another embodiment of the Taguchi method. Modern process control software is quite smart about adjusting manufacturing process for disturbances that it is pre-programmed to understand. Cognitive Plant Advisor plugs the gaps in the process control software's knowledge by learning how to adjust the process for fluctuations in variables that the process control software doesn't monitor.
Predictive Maintenance: My first cloud native SaaS App. In collaboration with a team of IBM researchers, we productized another embodiment of the Taguchi method. Modern process control software is quite smart about adjusting manufacturing process for disturbances that it is pre-programmed to understand. Cognitive Plant Advisor plugs the gaps in the process control software's knowledge by learning how to adjust the process for fluctuations in variables that the process control software doesn't monitor. Any machine. Any operating data. Maintenance recommendations.
Cloud Native. Architecture and technology are just a means to an end. My architectural process puts the business problem first - and the latest industry buzzwords and technology trends come a distant second. There is one exception. After using many (and building one) unsuccessful adaptation s of on-prem software to the cloud, my first cloud native app taught me that it is much easier to re-assemble apps on the cloud from existing cloud service than move anything that wasn't designed for the cloud into the cloud.
After working on several AI related apps, I established a design methodology for incorporating AI into software applications. It is very different from the typical data science approaches like CRISP-DM. The main difference is that I am more concerned with "decisions" than I am with "data". When designing a product using this methodology, the aim is to create a product that is easy to deploy and delivers quantifiable benefit - even if the customer has little data or poor data.
This methodology reduced the design effort for new products. Next, I set to work on reducing the implementation effort.
I conceived a way that would give IoT Device manufacturers a painless to offer their own software products for their devices. The IoT Analytic Service is a platform for delivering analysis. It is 100% automated and UI driven, and fully extensible. Customers can use predefined functions or develop their own functions in python. In some cases it reduced time to market for new applications to as little as 1 month.
As chief architect for a suite of IoT apps, the work I did was highly varied: I did a lot of hands on coding - mainly in the form of simulations that I used to test predictive models. I was part of the UX design process. During the ideation process for each app, the design team and I would iterate through a full range of concepts. I was active with field teams: trained at least 50 pre-sales and post-sales consultants and advised dozens of customers on what they might be able to do with the IoT data that they were collecting. I also got to present at conferences, serve on the board of a conference, and run public workshops. Panelist at the Cesis conference. Speaker at the VDI Maintenance conference. On board of VID AI conference. Workshop leader for the VDI Machine Learning for better Maintenance Decisions Conference.
I ran incubation projects that explored advances in the state of the art of business modeling, planning, AI-assisted application building, business simulation and decision automation.
To get in touch, email mike@adendorff.com. Site and Illustrations © Mike Adendorff - All rights reserved