Implement some checks based on abstract interpretation.
Enhance the graph backend to show the C source code and the linearized bytecode.
Use Sparse to generate FFI declarations for some higher-level language, such as Haskell, Python, or Java.
Do something to parse assembly code or object code and turn it into some approximation to the Sparse linearized instruction format. This will help with analysis of the Linux kernel.
Ideally, this would also support inline assembly blocks. Those may even prove easier to start on.
Write a compiler backend to generate code for some real architecture. Sparse already has a toy compiler backend for x86, but it handles only a small subset of C.
Use the Sparse abstract syntax tree or linearized instruction format to generate bytecode for Java or .NET CLI.