diff --git a/docs/developer-guide/architecture.md b/docs/developer-guide/architecture.md
new file mode 100644
index 000000000..33c0f3d49
--- /dev/null
+++ b/docs/developer-guide/architecture.md
@@ -0,0 +1,25 @@
+# Neural-LAM Architecture & Developer Guide
+
+
+## High-Level Overview
+
+Neural-LAM is a graph-based neural weather prediction system for Limited Area Modeling (LAM).
+The core pipeline is: **Data → Graph → Model → Training → Forecast**.
+
+```mermaid
+flowchart TD
+    A[Raw Input Data\n(xarray / MEPS / NpyFilesDatastore)]
+    --> B[Configuration & Datastore Loading\nload_config_and_datastore\nconfig.py]
+
+    B --> C[Graph Construction\ncreate_graph\nMulti-scale / Hierarchical\nRectangular or Triangular]
+
+    C --> D[Graph Neural Network Model\nGraphLAM or HiLAM\nMessage Passing on Graph]
+
+    D --> E[Autoregressive Forecasting\nAR steps\nEnsemble / Probabilistic Extensions\nLBCs handling]
+
+    E --> F[Loss & Optimization\nState feature weighting\nOutput clamping\nTraining loop]
+
+    F --> G[Output & Evaluation\nForecasts + Visualization]
+
+    style A fill:#e3f2fd,stroke:#1976d2
+    style G fill:#f3e5f5,stroke:#7b1fa2