Creating a new clustering algorithm with java and rapidminer

Hi, I'm new to RapidMiner. I have a project that consists in modeling a new clustering algorithm using the Java programming language. For example, my model starts by applying K-means in the first time and then add other techniques. In this case, I don't know what I should do. I researched on the net and I don't understand: do I  must  integrate RapidMiner in my java code then use the algorithms implemented in rapidminer even be able to see the representation as in RapidMiner (in my case clustering the data) to compare my algorithm with existing algorithms, or I must to model this new model in RapidMiner then add it to my java application.
I have integrated RapidMiner in Eclipse like is described in : http://rapid-i.com/content/view/25/48/lang,en/.
I didn't understand the utility of this, can I use for example the K-means algorithm from rapidMiner and recuperate the results in my java code ?
Things are mixed in my head. Can you help me to know how to start.
best regards.