Getting your AI project started
AI or ML is certainly the hottest topic in the business/IT world at the moment but what is the best way to get started and create something that is of value to the business?
Traditionally AI has meant Artificial Intelligence, but a more appropriate definition is Augmented Intelligence – providing the business with the mechanism to automate decision making. Most companies have some kind of dashboard that shows key business or process indicators but often these dashboards input into a manual decision making process. AI/ML provides the ability to automate decision making so that decisions are based on data rather than opinion.
So what is the best way to start your project or proof of concept? Obviously the scope needs to be well defined and the issue it is aimed at resolving. However also consider the type of data that will feed your analytics. For example, your data analysis could be based on:
- Business Process data such as your own corporate data and transactions
- Sensor based data such as satellite or from IoT devices
- Social Media or news/reviews data
Also, do you need any external feeds to complement your existing data pool?
Once you have an appropriate data set you can then decide (with the help of your data scientist or analyst) the appropriate open source framework to use – maybe something like Tensorflow, Caffe, Torch or one of the many that exist based on your requirements (ease of use, scalability, portability, availability of documentation etc).
If you don’t have a data scientist there are many companies that can provide assistance, however you probably do already have staff that analyse your current data and trends so you are not necessarily starting from zero.The next key thing is where to run your PoC? The big public cloud providers have environments with the necessary GPUs but these aren’t necessarily cost effective and can suffer from speed of processing and scalability issues. Alternatively, you can consider a platform based on IBM’s PowerAI (Power9) technology that has faster CPU<->GPU processing speeds than x86 based environments and can scale without performance being sacrificed. L3C (https://l3ccloud.com/ai-hosting/) provide access to such environments in a UK cloud model enabling you to cost effectively run your PoC and ultimately deploy on the latest AI oriented technology. We can also integrate with your legacy applications as well as those already deployed in public cloud.
For an informal conversation on your needs contact us at firstname.lastname@example.org