diff --git a/backend/services/caller_gen.py b/backend/services/caller_gen.py
new file mode 100644
index 0000000..be8c1e8
--- /dev/null
+++ b/backend/services/caller_gen.py
@@ -0,0 +1,38 @@
+from dataclasses import dataclass
+from typing import Optional
+import json
+
+REQUIRED_FIELDS = {
+    "name", "age", "voice_suggestion", "location", "identity",
+    "situation", "reason_calling", "opening_line", "secret_want",
+    "specific_details", "emotional_register"
+}
+
+
+@dataclass
+class CallerIdentity:
+    name: str
+    age: int
+    voice_suggestion: str
+    location: str
+    identity: str
+    situation: str
+    reason_calling: str
+    opening_line: str
+    secret_want: str
+    specific_details: list[str]
+    emotional_register: str
+    # set after voice validation
+    voice_resolved: Optional[str] = None
+
+
+def parse_batch_response(raw: str) -> list[CallerIdentity]:
+    data = json.loads(raw)
+    callers = data.get("callers", [])
+    result = []
+    for c in callers:
+        missing = REQUIRED_FIELDS - set(c.keys())
+        if missing:
+            raise ValueError(f"CallerIdentity missing fields: {missing}")
+        result.append(CallerIdentity(**{k: c[k] for k in REQUIRED_FIELDS}))
+    return result
diff --git a/tests/test_caller_gen.py b/tests/test_caller_gen.py
new file mode 100644
index 0000000..e49c94e
--- /dev/null
+++ b/tests/test_caller_gen.py
@@ -0,0 +1,37 @@
+import pytest
+from backend.services.caller_gen import CallerIdentity, parse_batch_response
+
+SAMPLE_JSON = """
+{
+  "callers": [
+    {
+      "name": "Danny Ortega",
+      "age": 47,
+      "voice_suggestion": "Marcus",
+      "location": "Silver City, NM",
+      "identity": "A plumber who inherited his uncle's taxidermy shop...",
+      "situation": "He's been getting calls from people...",
+      "reason_calling": "Someone left a note in his mailbox tonight...",
+      "opening_line": "Luke, I need to ask you something weird.",
+      "secret_want": "Permission to just throw it all away",
+      "specific_details": ["the elk head in the basement", "the note said 'she forgot'", "his uncle's Rolodex"],
+      "emotional_register": "quietly unsettled, trying to sound casual"
+    }
+  ]
+}
+"""
+
+
+def test_parse_batch_response_returns_caller_list():
+    callers = parse_batch_response(SAMPLE_JSON)
+    assert len(callers) == 1
+    assert callers[0].name == "Danny Ortega"
+    assert callers[0].age == 47
+    assert "taxidermy" in callers[0].identity
+    assert len(callers[0].specific_details) == 3
+
+
+def test_parse_batch_response_rejects_missing_fields():
+    bad = '{"callers": [{"name": "Jim"}]}'
+    with pytest.raises(ValueError, match="missing"):
+        parse_batch_response(bad)