diff --git a/Makefile b/Makefile
index 75fbb43..6cf070a 100644
--- a/Makefile
+++ b/Makefile
@@ -25,7 +25,7 @@ endef
 export PRINT_HELP_PYSCRIPT
 
 BROWSER := python -c "$$BROWSER_PYSCRIPT"
-PY_PATHS = ncollpyde tests
+PY_PATHS = python/ncollpyde tests
 
 help:
 	@python -c "$$PRINT_HELP_PYSCRIPT" < $(MAKEFILE_LIST)
@@ -87,7 +87,7 @@ install-opt:
 	maturin develop --release
 
 coverage: install-dev
-	coverage run --source ncollpyde -m pytest && \
+	coverage run --source python/ncollpyde -m pytest && \
 	coverage report -m && \
 	coverage html && \
 	$(BROWSER) htmlcov/index.html
diff --git a/pyproject.toml b/pyproject.toml
index 07a6d81..7798565 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -37,6 +37,10 @@ repository = "https://github.com/clbarnes/ncollpyde/"
 requires = ["maturin==0.14", "numpy>=1.21"]
 build-backend = "maturin"
 
+[tool.maturin]
+python-source = "python"
+module-name = "ncollpyde._ncollpyde"
+
 [tool.mypy]
 ignore_missing_imports = true
 plugins = ["numpy.typing.mypy_plugin"]
diff --git a/ncollpyde/__init__.py b/python/ncollpyde/__init__.py
similarity index 88%
rename from ncollpyde/__init__.py
rename to python/ncollpyde/__init__.py
index 16d9c0b..96614cc 100644
--- a/ncollpyde/__init__.py
+++ b/python/ncollpyde/__init__.py
@@ -13,8 +13,8 @@
 from .main import PRECISION  # noqa: F401
 from .main import Volume  # noqa: F401
 from .main import configure_threadpool  # noqa: F401
-from .ncollpyde import n_threads  # noqa: F401
-from .ncollpyde import _version
+from ._ncollpyde import n_threads  # noqa: F401
+from ._ncollpyde import _version
 
 __version__ = _version()
 __version_info__ = tuple(int(n) for n in __version__.split("-")[0].split("."))
diff --git a/ncollpyde/ncollpyde.pyi b/python/ncollpyde/_ncollpyde.pyi
similarity index 92%
rename from ncollpyde/ncollpyde.pyi
rename to python/ncollpyde/_ncollpyde.pyi
index d7fbb01..f94a936 100644
--- a/ncollpyde/ncollpyde.pyi
+++ b/python/ncollpyde/_ncollpyde.pyi
@@ -7,7 +7,7 @@ def _precision() -> str: ...
 def _index() -> str: ...
 def _version() -> str: ...
 def n_threads() -> int: ...
-def configure_threadpool(n_threads: Optional[int], name_prefix: Optional[str]): ...
+def _configure_threadpool(n_threads: Optional[int], name_prefix: Optional[str]): ...
 
 Points = npt.NDArray[np.float64]
 Indices = npt.NDArray[np.uint32]
diff --git a/ncollpyde/main.py b/python/ncollpyde/main.py
similarity index 99%
rename from ncollpyde/main.py
rename to python/ncollpyde/main.py
index f201927..0e651ba 100644
--- a/ncollpyde/main.py
+++ b/python/ncollpyde/main.py
@@ -12,11 +12,11 @@
 except ImportError:
     trimesh = None
 
-from .ncollpyde import (
+from ._ncollpyde import (
     TriMeshWrapper,
     _index,
     _precision,
-    configure_threadpool as _configure_threadpool,
+    _configure_threadpool,
 )
 
 if TYPE_CHECKING:
diff --git a/src/interface.rs b/src/interface.rs
index d8c8e1e..f3a5319 100644
--- a/src/interface.rs
+++ b/src/interface.rs
@@ -221,6 +221,7 @@ impl TriMeshWrapper {
 }
 
 #[pymodule]
+#[pyo3(name = "_ncollpyde")]
 pub fn ncollpyde(_py: Python, m: &PyModule) -> PyResult<()> {
     m.add_class::<TriMeshWrapper>()?;
 
@@ -249,7 +250,7 @@ pub fn ncollpyde(_py: Python, m: &PyModule) -> PyResult<()> {
     }
 
     #[pyfn(m)]
-    #[pyo3(name = "configure_threadpool")]
+    #[pyo3(name = "_configure_threadpool")]
     pub fn configure_threadpool(
         _py: Python,
         n_threads: Option<usize>,
diff --git a/tests/test_ncollpyde.py b/tests/test_ncollpyde.py
index c288641..743d143 100644
--- a/tests/test_ncollpyde.py
+++ b/tests/test_ncollpyde.py
@@ -5,6 +5,7 @@
 from itertools import product
 import sys
 import subprocess as sp
+import logging
 
 import numpy as np
 import pytest
@@ -16,6 +17,8 @@
 
 from ncollpyde import PRECISION, Volume, configure_threadpool
 
+logger = logging.getLogger(__name__)
+
 points_expected = [
     ([-2.3051376, -4.1556454, 1.9047838], True),  # internal
     ([-0.35222054, -0.513299, 7.6191354], False),  # external but in AABB
@@ -300,7 +303,10 @@ def test_configure_threadpool_subprocess():
     )
     args = [sys.executable, "-c", cmd]
 
-    sp.run(args, check=True, text=True, capture_output=True)
+    result = sp.run(args, text=True, capture_output=True)
+    logger.info(result.stdout)
+    logger.warning(result.stderr)
+    result.check_returncode()
 
 
 def test_configure_threadpool_twice():