diff --git a/iree/turbine/kernel/wave/scheduling/graph_utils.py b/iree/turbine/kernel/wave/scheduling/graph_utils.py
index e3d759dc..610c56cd 100644
--- a/iree/turbine/kernel/wave/scheduling/graph_utils.py
+++ b/iree/turbine/kernel/wave/scheduling/graph_utils.py
@@ -15,6 +15,7 @@
 from functools import partial
 from ..utils import safe_subs
 import multiprocessing as mp
+from typing import Optional
 
 T = index_symbol("$INITIATION_INTERVAL")
 
@@ -203,7 +204,7 @@ def all_pairs_longest_paths_symbolic(
 
 
 def all_pairs_longest_paths(
-    graph: fx.Graph, edges: list[Edge], T: int, pool: mp.Pool
+    graph: fx.Graph, edges: list[Edge], T: int, pool: Optional[mp.Pool]
 ) -> dict[tuple[fx.Node, fx.Node], IndexExpr]:
     """
     For each node in the graph, compute the longest path to all other nodes.
@@ -229,7 +230,11 @@ def all_pairs_longest_paths(
     # Parallel implementation
     for k in range(N):
         func = partial(all_pairs_longest_path_parallel, N, D, k)
-        results = pool.map(func, range(N))
+        if pool is not None:
+            results = pool.map(func, range(N))
+        else:
+            results = map(func, range(N))
+
         for result in results:
             D[result[0]] = result[1]
 
diff --git a/iree/turbine/kernel/wave/scheduling/modulo_scheduling.py b/iree/turbine/kernel/wave/scheduling/modulo_scheduling.py
index 7de08162..fb7794e4 100644
--- a/iree/turbine/kernel/wave/scheduling/modulo_scheduling.py
+++ b/iree/turbine/kernel/wave/scheduling/modulo_scheduling.py
@@ -111,7 +111,16 @@ def schedule_graph(self) -> tuple[dict[fx.Node, int], bool]:
         # TODO: Come up with a better heuristic on an upper bound for the initiation interval.
         T_max_range = 3 * T0
         success = False
-        pool = mp.get_context("fork").Pool(processes=mp.cpu_count())
+
+        # We cannot create create child processes when running in daemon process
+        # so just run sequentially.
+        # TODO: Find a way to reuse processes from the outside pool if we are
+        # already running inside.
+        if mp.current_process().daemon:
+            pool = None
+        else:
+            pool = mp.get_context("fork").Pool(processes=mp.cpu_count())
+
         for T in range(T0, T0 + T_max_range):
             logger.debug(f"Trying initiation interval: {T}.")
             self.RT = np.zeros((T, len(self.resources)))
@@ -150,8 +159,10 @@ def schedule_graph(self) -> tuple[dict[fx.Node, int], bool]:
                 break
         else:
             raise Exception("Failed to schedule the graph.")
-        pool.close()
-        pool.join()
+
+        if pool is not None:
+            pool.close()
+            pool.join()
 
         self._initiation_interval = T
         return self.schedule, success