diff --git a/services/ml_training_data/pyproject.toml b/services/ml_training_data/pyproject.toml
index 15e00384..12b412a5 100644
--- a/services/ml_training_data/pyproject.toml
+++ b/services/ml_training_data/pyproject.toml
@@ -11,6 +11,8 @@ dependencies = [
     "pyarrow>=15",
     "lightgbm>=4.0",
     "scikit-learn>=1.4",
+    "httpx",
+    "remotezip>=0.12",
 ]
 
 [tool.uv.sources]
diff --git a/services/ml_training_data/src/ml_training_data/remote_bulk_fetcher.py b/services/ml_training_data/src/ml_training_data/remote_bulk_fetcher.py
new file mode 100644
index 00000000..afa16263
--- /dev/null
+++ b/services/ml_training_data/src/ml_training_data/remote_bulk_fetcher.py
@@ -0,0 +1,79 @@
+"""Extract specific yearly entries from the gov bulk JSON ZIP without downloading
+the whole 15 GB archive.
+
+The gov endpoint returns a 302 to a pre-signed S3 URL. remotezip uses HTTP Range
+requests against that URL to read only the central directory + the bytes for the
+requested entries, so disk usage stays at "size of the entries we actually want"
+instead of the full archive.
+
+Entries are streamed via zipfile.ZipExtFile.read(chunk) so partial-network failures
+during the multi-GB read don't waste the whole transfer, and so we never hold the
+full entry in memory.
+"""
+
+from pathlib import Path
+from tempfile import NamedTemporaryFile
+
+import httpx
+from remotezip import RemoteZip  # type: ignore[import-untyped]  # pyright: ignore[reportMissingTypeStubs]
+
+from ml_training_data.storage import Storage
+
+_BULK_JSON_URL = (
+    "https://api.get-energy-performance-data.communities.gov.uk/api/files/domestic/json"
+)
+_READ_CHUNK_BYTES = 8 * 1024 * 1024  # 8 MB
+
+
+def extract_entries(
+    auth_token: str,
+    entry_names: list[str],
+    storage: Storage,
+    key_prefix: str,
+) -> dict[str, int]:
+    presigned_url = _resolve_presigned_url(auth_token)
+    sizes: dict[str, int] = {}
+    with RemoteZip(presigned_url) as zf:  # pyright: ignore[reportUnknownVariableType]
+        for entry in entry_names:
+            n_bytes = _stream_entry_to_storage(zf, entry, storage, f"{key_prefix}{entry}")
+            sizes[entry] = n_bytes
+    return sizes
+
+
+def _stream_entry_to_storage(
+    zf: RemoteZip,  # pyright: ignore[reportUnknownParameterType]
+    entry: str,
+    storage: Storage,
+    output_key: str,
+) -> int:
+    with NamedTemporaryFile(delete=False) as tmp:
+        tmp_path = Path(tmp.name)
+        with zf.open(entry) as src:  # pyright: ignore[reportUnknownMemberType,reportUnknownVariableType]
+            while True:
+                chunk: bytes = src.read(_READ_CHUNK_BYTES)  # pyright: ignore[reportUnknownVariableType,reportUnknownMemberType]
+                if not chunk:
+                    break
+                tmp.write(chunk)
+        total = tmp_path.stat().st_size
+    storage.write_bytes(output_key, tmp_path.read_bytes())
+    tmp_path.unlink()
+    return total
+
+
+def _resolve_presigned_url(auth_token: str) -> str:
+    response = httpx.get(
+        _BULK_JSON_URL,
+        headers={"Authorization": f"Bearer {auth_token}"},
+        follow_redirects=False,
+        timeout=30,
+    )
+    if response.status_code != 302:
+        raise RuntimeError(
+            f"Bulk JSON endpoint did not redirect: {response.status_code} {response.text[:200]}"
+        )
+    location = response.headers.get("location")
+    if not location:
+        raise RuntimeError("Bulk JSON 302 had no Location header")
+    return location
+
+