Although a growing number of devices have the support for parallel processing, parallelism is not widely exploited, as it should be. This can be illustrated by the fact that all Apple desktops and laptops are currently supplied with one or more dual-core processors. Despite this, parallel programming in popular languages such as Java is not widely encouraged and often only recommended as a last resort. In addition, it is likely that the next generation parallel applications will have to operate within a diverse range of heterogeneous devices ranging from supercomputers to sensors. This paper proposes a flexible component-based middleware that aims at facilitating the construction of deadlock-free parallel and concurrent applications for heterogeneous environments. The middleware is particularly targeted to tailor applications to the target need and environment. For our prototyping, we implemented plugins that are capable of deploying JCSP(CSP library for Java programmers) and occampi processes. Both JCSP and occampi make use of the CSP disciplines. The CSP (Communicating Sequential Processes) paradigm helps us to avoid concurrency problems such as deadlocks. It should be stressed that there is no bias towards these languages as the proposed middleware is highly extensible.