diff --git a/gantry/routes/prediction.py b/gantry/routes/prediction.py
index 7ef8c4f..5f000fc 100644
--- a/gantry/routes/prediction.py
+++ b/gantry/routes/prediction.py
@@ -12,6 +12,7 @@
 DEFAULT_CPU_LIMIT = 5
 DEFAULT_MEM_LIMIT = 5 * 1_000_000_000
 MEM_LIMIT_BUMP = 1.2
+MEMORY_LIMIT_FLOOR = 3.5 * 1_000_000  # 350MB
 EXPENSIVE_VARIANTS = {
     "sycl",
     "mpi",
@@ -44,7 +45,7 @@ async def predict(db: aiosqlite.Connection, spec: dict) -> dict:
             "mem_request": DEFAULT_MEM_REQUEST,
             "cpu_limit": DEFAULT_CPU_LIMIT,
             "mem_limit": DEFAULT_MEM_LIMIT,
-            "build_jobs": DEFAULT_CPU_REQUEST,
+            "build_jobs": DEFAULT_CPU_LIMIT,
         }
     else:
         # mapping of sample: [0] cpu_mean, [1] cpu_max, [2] mem_mean, [3] mem_max
@@ -56,7 +57,9 @@ async def predict(db: aiosqlite.Connection, spec: dict) -> dict:
             "mem_limit": max(build[3] for build in sample) * MEM_LIMIT_BUMP,
         }
         # build jobs cannot be less than 1
-        predictions["build_jobs"] = max(1, round(predictions["cpu_request"]))
+        predictions["build_jobs"] = max(1, round(predictions["cpu_limit"]))
+        # enforce memory limit floor
+        predictions["mem_limit"] = max(predictions["mem_limit"], MEMORY_LIMIT_FLOOR)
 
     # convert predictions to k8s friendly format
     for k, v in predictions.items():
diff --git a/gantry/tests/defs/prediction.py b/gantry/tests/defs/prediction.py
index 7ba2624..e567b46 100644
--- a/gantry/tests/defs/prediction.py
+++ b/gantry/tests/defs/prediction.py
@@ -39,6 +39,6 @@
         "KUBERNETES_CPU_REQUEST": "1000m",
         "KUBERNETES_MEMORY_LIMIT": "5000M",
         "KUBERNETES_MEMORY_REQUEST": "2000M",
-        "SPACK_BUILD_JOBS": 1,
+        "SPACK_BUILD_JOBS": 5,
     },
 }