diff --git a/viscy/data/gpu_aug.py b/viscy/data/gpu_aug.py
index a445fff5..09302aad 100644
--- a/viscy/data/gpu_aug.py
+++ b/viscy/data/gpu_aug.py
@@ -197,6 +197,8 @@ class CachedOmeZarrDataModule(GPUTransformDataModule):
         Use page-locked memory in data-loaders, by default True
     skip_cache : bool, optional
         Skip caching for this dataset, by default False
+    include_wells : list[str], optional
+        List of well names to include in the dataset, by default None (all)
     """
 
     def __init__(
@@ -212,6 +214,7 @@ def __init__(
         val_gpu_transforms: list[DictTransform],
         pin_memory: bool = True,
         skip_cache: bool = False,
+        include_wells: list[str] | None = None,
     ):
         super().__init__()
         self.data_path = data_path
@@ -225,6 +228,7 @@ def __init__(
         self._val_gpu_transforms = Compose(val_gpu_transforms)
         self.pin_memory = pin_memory
         self.skip_cache = skip_cache
+        self._include_wells = include_wells
 
     @property
     def train_cpu_transforms(self) -> Compose:
@@ -248,12 +252,30 @@ def _set_fit_global_state(self, num_positions: int) -> list[int]:
         # shuffle positions, randomness is handled globally
         return torch.randperm(num_positions).tolist()
 
+    def _include_well_name(self, name: str) -> bool:
+        if self._include_wells is None:
+            return True
+        else:
+            return name in self._include_wells
+
+    def _filter_fit_fovs(self, plate: Plate) -> list[Position]:
+        positions = []
+        for well_name, well in plate.wells():
+            if self._include_well_name(well_name):
+                for _, p in well.positions():
+                    positions.append(p)
+        if len(positions) < 2:
+            raise ValueError(
+                "At least 2 FOVs are required for training and validation."
+            )
+        return positions
+
     def setup(self, stage: Literal["fit", "validate"]) -> None:
         if stage not in ("fit", "validate"):
             raise NotImplementedError("Only fit and validate stages are supported.")
         cache_map = Manager().dict()
         plate: Plate = open_ome_zarr(self.data_path, mode="r", layout="hcs")
-        positions = [p for _, p in plate.positions()]
+        positions = self._filter_fit_fovs(plate)
         shuffled_indices = self._set_fit_global_state(len(positions))
         num_train_fovs = int(len(positions) * self.split_ratio)
         train_fovs = [positions[i] for i in shuffled_indices[:num_train_fovs]]