diff --git a/apis/python/anndata/subset-soma-01.h5ad b/apis/python/anndata/subset-soma-01.h5ad
new file mode 100644
index 0000000000..d2ec6322ec
Binary files /dev/null and b/apis/python/anndata/subset-soma-01.h5ad differ
diff --git a/apis/python/anndata/subset-soma-02.h5ad b/apis/python/anndata/subset-soma-02.h5ad
new file mode 100644
index 0000000000..5e568b36ed
Binary files /dev/null and b/apis/python/anndata/subset-soma-02.h5ad differ
diff --git a/apis/python/anndata/subset-soma-03.h5ad b/apis/python/anndata/subset-soma-03.h5ad
new file mode 100644
index 0000000000..778816d628
Binary files /dev/null and b/apis/python/anndata/subset-soma-03.h5ad differ
diff --git a/apis/python/anndata/subset-soma-04.h5ad b/apis/python/anndata/subset-soma-04.h5ad
new file mode 100644
index 0000000000..d5c4b30f79
Binary files /dev/null and b/apis/python/anndata/subset-soma-04.h5ad differ
diff --git a/apis/python/tools/generate-small-anndata.py b/apis/python/tools/generate-small-anndata
similarity index 100%
rename from apis/python/tools/generate-small-anndata.py
rename to apis/python/tools/generate-small-anndata
diff --git a/apis/python/tools/splitter b/apis/python/tools/splitter
new file mode 100644
index 0000000000..c5dfc21f5f
--- /dev/null
+++ b/apis/python/tools/splitter
@@ -0,0 +1,47 @@
+#!/usr/bin/env python
+
+"""
+Given one of the outputs from cartorapher.py, splits it into a few little pieces for quick-iteration purposes.
+"""
+
+import tiledbsc
+import tiledbsc.io
+
+import numpy as np
+import os, shutil
+
+# ----------------------------------------------------------------
+def write_subset(soma, obs_indices, output_path):
+    if os.path.exists(output_path):
+        shutil.rmtree(output_path)
+    subset_soma = tiledbsc.SOMA(output_path)
+    subset_soma.create_unless_exists()
+
+    obs_ids = list(soma.obs.df().index[obs_indices])
+    var_ids = list(soma.var.df().index)
+
+    subset_obs = soma.obs.df(obs_ids)
+    subset_var = soma.var.df()
+    subset_X_data = soma.X.data.csr(obs_ids, None)
+
+    subset_obs["is_primary_data"] = np.asarray([True] * len(obs_ids))
+
+    subset_soma.obs.from_dataframe(subset_obs, extent=2048)
+    subset_soma.var.from_dataframe(subset_var, extent=2048)
+
+    print("S OBS", subset_soma.obs.shape())
+    print("S VAR", subset_soma.var.shape())
+    print("S XDA", subset_X_data.shape)
+
+    subset_soma.X.add_layer_from_matrix_and_dim_values(subset_X_data, obs_ids, var_ids)
+
+    tiledbsc.io.to_h5ad(subset_soma, output_path + ".h5ad")
+
+
+# ----------------------------------------------------------------
+input_soma = tiledbsc.SOMA("atlas/4056cbab-2a32-4c9e-a55f-c930bc793fb6")
+
+write_subset(input_soma, range(0, 100), "subset-soma-01")
+write_subset(input_soma, range(100, 200), "subset-soma-02")
+write_subset(input_soma, range(200, 300), "subset-soma-03")
+write_subset(input_soma, range(300, 400), "subset-soma-04")