The recommended build-from-source method is to use pip packaging. If, however, you're running into issues, you can manually build intel-qs using Anaconda and deal with module sorcery yourself.
The general flow for Unix is to make sure you have python-dev(aka python-devel), make and g++. This is, of course, in addition to python and pip, which you hopefully already have. Keep in mind some of these tools may come pre-bundled with others. If in doubt, which make, which g++, etc. to verify.
Below are verified-correct setups to builds from source of the pip package. Once you've done the "prep", you should be able to just run pip install . in this folder and build/install the package. You also need cmake and pybind11 to build intel-qs but they should be pulled by pip tooling automatically.
# You can skip git/python/pip setup if you have all of them
sudo yum install git
sudo yum install python3
curl https://bootstrap.pypa.io/pip/3.6/get-pip.py > get-pip.py
python3 get-pip.py
sudo yum install make
sudo yum install platform-python-devel.x86_64
pip install .Building manually is probably something you should only try if pip is giving you trouble (and you should report that on GitHub Issues). You will need to install Anaconda to proceed. Verify successful installation by checking that which python returns an anaconda path prior to attempting any intel-qs setup.
If you're on Unix (including OS X), you should be able to get going by running manual-setup.sh and waiting a while. If you're on Windows, you can still do that by using WSL (highly recommended).
The end result of running manual-setup.sh is that you (hopefully) now have intel-qs built from the Python/C++ source. However, the qublets package is not connected anywhere, since you didn't use pip. You can copy the contents of intel-qs/build/lib and qublets (preserving the folder structure) anywhere you'd like to have the qublets module accessible and handle module dependencies manually.