diff --git a/src/Components/API/app/routers/species_predictor.py b/src/Components/API/app/routers/species_predictor.py
index 81e8373ce..ef305408c 100644
--- a/src/Components/API/app/routers/species_predictor.py
+++ b/src/Components/API/app/routers/species_predictor.py
@@ -2,18 +2,12 @@
 from fastapi.responses import JSONResponse
 from typing import Optional
 from app.database import Predictions
+from app.services.model_adapter import MultiModalPredictionError, predict_with_failure_detection
 import datetime
 import re
 
 router = APIRouter()
 
-# Placeholder prediction
-def predict_species(audio_file: UploadFile):
-    return {
-        "species": "Crimson Rosella",
-        "confidence": 0.92
-    }
-
 @router.post("/predict")
 async def predict(
     audio: UploadFile = File(...),
@@ -31,7 +25,10 @@ async def predict(
         else:
             raise HTTPException(status_code=400, detail="Invalid upload_id format")
 
-    prediction = predict_species(audio)
+    try:
+        prediction = predict_with_failure_detection(audio)
+    except MultiModalPredictionError as e:
+        raise HTTPException(status_code=502, detail=str(e))
 
     # Persist prediction result
     try:
diff --git a/src/Components/API/app/services/model_adapter.py b/src/Components/API/app/services/model_adapter.py
new file mode 100644
index 000000000..2473c7f61
--- /dev/null
+++ b/src/Components/API/app/services/model_adapter.py
@@ -0,0 +1,44 @@
+from numbers import Number
+from typing import Any, Dict
+
+
+class MultiModalPredictionError(Exception):
+    """Raised when the multi-modal prediction path fails or returns invalid data."""
+
+
+def predict_multimodal(audio_file: Any) -> Dict[str, Any]:
+    # Placeholder until the real multi-modal engine is wired into the adapter.
+    return {
+        "species": "Crimson Rosella",
+        "confidence": 0.92,
+    }
+
+
+def validate_prediction_response(result: Any) -> Dict[str, Any]:
+    if not isinstance(result, dict):
+        raise MultiModalPredictionError("Multi-modal model returned an invalid response type")
+
+    species = result.get("species")
+    confidence = result.get("confidence")
+
+    if not isinstance(species, str) or not species.strip():
+        raise MultiModalPredictionError("Multi-modal model response is missing a valid species")
+
+    if not isinstance(confidence, Number):
+        raise MultiModalPredictionError("Multi-modal model response is missing a valid confidence")
+
+    return {
+        "species": species,
+        "confidence": float(confidence),
+    }
+
+
+def predict_with_failure_detection(audio_file: Any) -> Dict[str, Any]:
+    try:
+        result = predict_multimodal(audio_file)
+    except MultiModalPredictionError:
+        raise
+    except Exception as exc:
+        raise MultiModalPredictionError("Multi-modal prediction failed") from exc
+
+    return validate_prediction_response(result)