Current rules-driven System is very cumbersome to maintain and is not able to keep up with new kinds of suspicious activity that is continuously evolving.
Machine Learning System that ingested millions of customer transactions over a 3-year time period was used to analyze the suspicious activity. A combination of Graph GAN + NEAT was used to build a predictive model for suspicious activity
Analyzes every transaction for suspicious activity and is able to flag suspicious behaviour at near real time
Dataiku, Tensorflow, Python, Apache Spark
Declining revenues from game sales in different regions in North America.
Built an AI Engine that analyzed data over a 4-year time period and was able to not only forecast the sales from games but also provide recommendations on what game types should be released in what regions and at what point of time.
An AI driven assistant for game planning was able to provide a 10% Lift to revenues in the next two quarter since it was operationalized.
Microsoft Azure, Tensorflow, Python, Things Stack network server, Tektelic sensors/ gateway, Azure SQL Server
Low income communities in municipalities in Canada that spend more than 10% of their disposable income towards meeting their energy needs - Need to reduce the spend on heating whilst maintaining a comfortable quality of life.
Energy retrofits and a data driven approach to measuring energy usage. Personalized recommendations on reducing energy consumptions
More disposable income for a better quality of life.
Dataiku, Tensorflow, Python, Apache Spark
