diff --git a/README.md b/README.md
new file mode 100644
index 0000000..e174374
--- /dev/null
+++ b/README.md
@@ -0,0 +1,52 @@
+# Operator Runtime Training Camp
+
+This repository is a training-oriented GPU operator runtime. It is intentionally
+small, but its workflow mirrors production operator libraries:
+
+1. Define the operator contract in `ops/<op>/operator.yaml`.
+2. Generate build and runtime registries from the manifest.
+3. Implement a backend-specific descriptor lifecycle.
+4. Expose a Python API with out-of-place, out-variant, and prepared execution.
+5. Validate correctness against PyTorch and benchmark steady-state execution.
+
+The project is rooted directly at `/workspace`; it does not create a nested
+`/workspace/camp` project directory.
+
+## Operators
+
+| Operator | NVIDIA C++ | TileLang | MetaX |
+| --- | --- | --- | --- |
+| `copy` | runnable | runnable when TileLang is installed | stub |
+| `vector_add` | runnable | runnable when TileLang is installed | stub |
+| `reduce_sum` | runnable, row-wise fp32 | runnable when TileLang is installed | stub |
+| `softmax` | runnable, row-wise fp32 | runnable when TileLang is installed | stub |
+
+## Build
+
+```bash
+mkdir -p build
+cd build
+cmake .. -DCAMP_ENABLE_NVIDIA=ON -DCAMP_ENABLE_METAX=OFF
+cmake --build . -j$(nproc)
+```
+
+## Validate
+
+```bash
+python tools/validate_operator_manifest.py --ops-root ops --tests-root tests
+CAMP_BUILD_DIR=build pytest tests/ -v --backend nvidia
+pytest tests/ -v --backend tilelang
+python tests/bench_all.py --backend nvidia --profile tests/perf_profiles/local_gpu.yaml
+```
+
+The TileLang backend requires the `tilelang` Python package. 
+
+## Production Mapping
+
+| Training concept | Production equivalent |
+| --- | --- |
+| `operator.yaml` | reviewed operator spec / manifest |
+| descriptor lifecycle | create, workspace, execute, destroy |
+| generated registry | operation table / backend registry |
+| `tests/cases/<op>.py` | correctness, layout, and API contract coverage |
+| `PerformanceResult` | profiler report row with latency, bytes, flops, bandwidth |
diff --git a/examples/01_copy.py b/examples/01_copy.py
new file mode 100644
index 0000000..4f8c8bf
--- /dev/null
+++ b/examples/01_copy.py
@@ -0,0 +1,23 @@
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT / "python"))
+
+import torch
+
+from operator_runtime import copy
+
+
+def main() -> None:
+    src = torch.randn((1024,), device="cuda", dtype=torch.float32)
+    out = copy(src, backend="nvidia")
+    torch.testing.assert_close(out, src)
+    print("copy ok")
+
+
+if __name__ == "__main__":
+    main()
+
diff --git a/examples/02_vector_add.py b/examples/02_vector_add.py
new file mode 100644
index 0000000..48c2685
--- /dev/null
+++ b/examples/02_vector_add.py
@@ -0,0 +1,24 @@
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT / "python"))
+
+import torch
+
+from operator_runtime import vector_add
+
+
+def main() -> None:
+    a = torch.randn((1024,), device="cuda", dtype=torch.float32)
+    b = torch.randn_like(a)
+    out = vector_add(a, b, backend="nvidia")
+    torch.testing.assert_close(out, a + b)
+    print("vector_add ok")
+
+
+if __name__ == "__main__":
+    main()
+
diff --git a/examples/03_reduce_sum.py b/examples/03_reduce_sum.py
new file mode 100644
index 0000000..2fcbc2e
--- /dev/null
+++ b/examples/03_reduce_sum.py
@@ -0,0 +1,23 @@
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT / "python"))
+
+import torch
+
+from operator_runtime import reduce_sum
+
+
+def main() -> None:
+    src = torch.randn((32, 128), device="cuda", dtype=torch.float32)
+    out = reduce_sum(src, dim=1, backend="nvidia")
+    torch.testing.assert_close(out, torch.sum(src, dim=1), atol=1e-5, rtol=1e-5)
+    print("reduce_sum ok")
+
+
+if __name__ == "__main__":
+    main()
+
diff --git a/examples/04_softmax.py b/examples/04_softmax.py
new file mode 100644
index 0000000..0ce1471
--- /dev/null
+++ b/examples/04_softmax.py
@@ -0,0 +1,23 @@
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT / "python"))
+
+import torch
+
+from operator_runtime import softmax
+
+
+def main() -> None:
+    src = torch.randn((32, 128), device="cuda", dtype=torch.float32)
+    out = softmax(src, dim=1, backend="nvidia")
+    torch.testing.assert_close(out, torch.softmax(src, dim=1), atol=1e-5, rtol=1e-5)
+    print("softmax ok")
+
+
+if __name__ == "__main__":
+    main()
+