diff --git a/ablate/core/__init__.py b/ablate/core/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ablate/core/cli/main.py b/ablate/core/cli/main.py
new file mode 100644
index 0000000..81495e1
--- /dev/null
+++ b/ablate/core/cli/main.py
@@ -0,0 +1,2 @@
+def main() -> None:  # TODO
+    pass
diff --git a/ablate/core/types/__init__.py b/ablate/core/types/__init__.py
new file mode 100644
index 0000000..f4e9ef1
--- /dev/null
+++ b/ablate/core/types/__init__.py
@@ -0,0 +1,4 @@
+from .runs import GroupedRun, Run
+
+
+__all__ = ["GroupedRun", "Run"]
diff --git a/ablate/core/types/runs.py b/ablate/core/types/runs.py
new file mode 100644
index 0000000..bf1c459
--- /dev/null
+++ b/ablate/core/types/runs.py
@@ -0,0 +1,33 @@
+from typing import Any, Dict, List, Tuple
+
+from pydantic import BaseModel
+
+
+class Run(BaseModel):
+    """A single run of an experiment.
+
+    Args:
+        id: Unique identifier for the run.
+        params: Parameters used for the run.
+        metrics: Metrics recorded during the run.
+        temporal: Temporal data recorded during the run.
+    """
+
+    id: str
+    params: Dict[str, Any]
+    metrics: Dict[str, float]
+    temporal: Dict[str, list[Tuple[int, float]]] = {}
+
+
+class GroupedRun(BaseModel):
+    """A collection of runs grouped by a key-value pair.
+
+    Args:
+        key: Key used to group the runs.
+        value: Value used to group the runs.
+        runs: List of runs that belong to this group.
+    """
+
+    key: str
+    value: str
+    runs: List[Run]
diff --git a/tests/core/test_types.py b/tests/core/test_types.py
new file mode 100644
index 0000000..b789406
--- /dev/null
+++ b/tests/core/test_types.py
@@ -0,0 +1,56 @@
+from pydantic import ValidationError
+import pytest
+
+from ablate.core.types import GroupedRun, Run
+
+
+def test_run() -> None:
+    run = Run(id="test_run", params={"param1": 1}, metrics={"metric1": 0.5})
+    assert run.id == "test_run"
+    assert run.params == {"param1": 1}
+    assert run.metrics == {"metric1": 0.5}
+    assert run.temporal == {}
+
+
+def test_temporal_data() -> None:
+    run = Run(
+        id="test_run",
+        params={"param1": 1},
+        metrics={"metric1": 0.5},
+        temporal={"metric1": [(0, 0.0), (1, 1.0), (2, 2.0)]},
+    )
+    assert run.temporal == {"metric1": [(0, 0.0), (1, 1.0), (2, 2.0)]}
+    assert run.temporal["metric1"][0] == (0, 0.0)
+
+
+def test_invalid_metrics_type() -> None:
+    with pytest.raises(ValidationError):
+        Run(id="bad", params={}, metrics="not a dict")
+
+
+def test_grouped_run() -> None:
+    run1 = Run(id="run1", params={"param1": 1}, metrics={"metric1": 0.5})
+    run2 = Run(id="run2", params={"param2": 2}, metrics={"metric2": 0.8})
+    grouped_run = GroupedRun(key="group_key", value="group_value", runs=[run1, run2])
+    assert grouped_run.key == "group_key"
+    assert grouped_run.value == "group_value"
+    assert len(grouped_run.runs) == 2
+    assert grouped_run.runs[0].id == "run1"
+    assert grouped_run.runs[1].id == "run2"
+
+
+def test_invalid_runs_type() -> None:
+    with pytest.raises(ValidationError):
+        GroupedRun(key="group_key", value="group_value", runs=["not a run"])
+
+
+def test_run_roundtrip_serialization() -> None:
+    run = Run(
+        id="run1",
+        params={"x": 1},
+        metrics={"acc": 0.9},
+        temporal={"loss": [(0, 0.1), (1, 0.05)]},
+    )
+    data = run.model_dump()
+    recovered = Run(**data)
+    assert recovered == run