diff --git a/tensorflow_probability/python/experimental/autobnn/BUILD b/tensorflow_probability/python/experimental/autobnn/BUILD
index d74d59673f..051f58b2ed 100644
--- a/tensorflow_probability/python/experimental/autobnn/BUILD
+++ b/tensorflow_probability/python/experimental/autobnn/BUILD
@@ -24,6 +24,23 @@ package(
     default_visibility = ["//visibility:public"],
 )
 
+py_library(
+    name = "autobnn.jax",
+    srcs = ["__init__.py"],
+    deps = [
+        ":bnn",
+        ":bnn_tree",
+        ":estimators",
+        ":kernels",
+        ":likelihoods",
+        ":models",
+        ":operators",
+        ":training_util",
+        ":util",
+        "//tensorflow_probability/python/internal:all_util",
+    ],
+)
+
 py_library(
     name = "bnn",
     srcs = ["bnn.py"],
@@ -97,6 +114,8 @@ py_library(
 py_test(
     name = "estimators_test",
     srcs = ["estimators_test.py"],
+    # TODO(b/322864412): enable this test in OSS.
+    tags = ["no-oss-ci"],
     deps = [
         ":estimators",
         ":kernels",
@@ -141,7 +160,6 @@ py_library(
         # flax:core dep,
         # jax dep,
         # jaxtyping dep,
-        "//tensorflow_probability:jax",
         "//tensorflow_probability/python/bijectors:softplus.jax",
         "//tensorflow_probability/python/distributions:distribution.jax",
         "//tensorflow_probability/python/distributions:inflated.jax",
@@ -200,7 +218,6 @@ py_library(
         ":likelihoods",
         # flax:core dep,
         # jax dep,
-        "//tensorflow_probability:jax",
         "//tensorflow_probability/python/bijectors:chain.jax",
         "//tensorflow_probability/python/bijectors:scale.jax",
         "//tensorflow_probability/python/bijectors:shift.jax",
@@ -238,7 +255,6 @@ py_library(
         # matplotlib dep,
         # numpy dep,
         # pandas dep,
-        "//tensorflow_probability:jax",
         "//tensorflow_probability/python/experimental/autobnn:bnn",
         "//tensorflow_probability/python/experimental/autobnn:util",
         "//tensorflow_probability/python/experimental/timeseries:metrics",
@@ -248,6 +264,8 @@ py_library(
 py_test(
     name = "training_util_test",
     srcs = ["training_util_test.py"],
+    # TODO(b/322864412): enable this test in OSS.
+    tags = ["no-oss-ci"],
     deps = [
         ":training_util",
         # chex dep,
diff --git a/tensorflow_probability/python/experimental/autobnn/__init__.py b/tensorflow_probability/python/experimental/autobnn/__init__.py
new file mode 100644
index 0000000000..d7e10a3d11
--- /dev/null
+++ b/tensorflow_probability/python/experimental/autobnn/__init__.py
@@ -0,0 +1,45 @@
+# Copyright 2024 The TensorFlow Probability Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ============================================================================
+"""Package for training GP-like Bayesian Neural Nets w/ composite structure."""
+
+from tensorflow_probability.python.experimental.autobnn import bnn
+from tensorflow_probability.python.experimental.autobnn import bnn_tree
+# estimators causes vectorized_stochastic_volatility_test to fail
+# because it imports training_util
+from tensorflow_probability.python.experimental.autobnn import estimators
+from tensorflow_probability.python.experimental.autobnn import kernels
+from tensorflow_probability.python.experimental.autobnn import likelihoods
+from tensorflow_probability.python.experimental.autobnn import models
+from tensorflow_probability.python.experimental.autobnn import operators
+# training_util causes vectorized_stochastic_volatility_test to fail
+# Suspects: JaxTyping, bayeux, matplotlib, pandas.
+# And the culprit is ... bayeux.
+from tensorflow_probability.python.experimental.autobnn import training_util
+from tensorflow_probability.python.experimental.autobnn import util
+from tensorflow_probability.python.internal import all_util
+
+_allowed_symbols = [
+    'bnn',
+    'bnn_tree',
+    'estimators',
+    'kernels',
+    'likelihoods',
+    'models',
+    'operators',
+    'training_util',
+    'util',
+]
+
+all_util.remove_undocumented(__name__, _allowed_symbols)
diff --git a/tensorflow_probability/python/experimental/autobnn/training_util.py b/tensorflow_probability/python/experimental/autobnn/training_util.py
index 20cc5da31f..a1d9d48663 100644
--- a/tensorflow_probability/python/experimental/autobnn/training_util.py
+++ b/tensorflow_probability/python/experimental/autobnn/training_util.py
@@ -27,10 +27,6 @@
 from tensorflow_probability.python.experimental.autobnn import bnn
 from tensorflow_probability.python.experimental.autobnn import util
 from tensorflow_probability.python.experimental.timeseries import metrics
-import tensorflow_probability.substrates.jax as tfp
-
-tfd = tfp.distributions
-tfb = tfp.bijectors
 
 
 def _make_bayeux_model(