diff --git a/projects/improve_sharedmem_when_multiprocessing_in_nodepipeline/README.md b/projects/improve_sharedmem_when_multiprocessing_in_nodepipeline/README.md
new file mode 100644
index 0000000..ac601d0
--- /dev/null
+++ b/projects/improve_sharedmem_when_multiprocessing_in_nodepipeline/README.md
@@ -0,0 +1,45 @@
+# Project title:
+
+### Key Investigators
+
+* Samuel Garcia
+* and more
+
+## Project Description
+
+Spikeinterface use a lot ProcessPoolExecutor which can imply buffer copy for multiprocessing.
+The idea is to use more systematically sharedmem buffer before the spawning.
+
+### Background
+
+Fill in with brief description and optional references [^1]
+
+### Objectives
+
+Fill in with brief objective for the project
+
+
+### Approach and Plan
+
+Fill in with approach and project plan
+
+ * [ ] Task 1
+ * [ ] Task 2
+ * [ ] Task 3
+
+### Progress
+
+Fill in with tentative milestones (or leave blacnk for later):
+
+ * [ ] Milestone 1
+ * [ ] Milestone 2
+ * [ ] Make PR to SpikeInterface
+
+### Next Steps (optional)
+
+Fill in with next steps
+
+## References
+
+[^1]: 
+
diff --git a/projects/improve_sharedmem_when_multiprocessing_in_nodepipeline/src/README.md b/projects/improve_sharedmem_when_multiprocessing_in_nodepipeline/src/README.md
new file mode 100644
index 0000000..80ea5e8
--- /dev/null
+++ b/projects/improve_sharedmem_when_multiprocessing_in_nodepipeline/src/README.md
@@ -0,0 +1 @@
+Add your code here