diff --git a/lucidia_llm/data/build_rlhf_pairs.py b/lucidia_llm/data/build_rlhf_pairs.py
new file mode 100644
index 0000000..4a9c47e
--- /dev/null
+++ b/lucidia_llm/data/build_rlhf_pairs.py
@@ -0,0 +1,6 @@
+"""Dataset builder for RLHF pairs."""
+
+
+def build_rlhf_pairs(input_path, output_path):
+    """Build RLHF training pairs from raw data."""
+    raise NotImplementedError("RLHF pairs builder not implemented")