diff --git a/inference.py b/inference.py
index 9fa8f24e..f6edde21 100644
--- a/inference.py
+++ b/inference.py
@@ -341,15 +341,6 @@ def inference_monomer_model(args):
     print("running in monomer mode...")
     config = model_config(args.model_name)
 
-    template_featurizer = templates.TemplateHitFeaturizer(
-        mmcif_dir=args.template_mmcif_dir,
-        max_template_date=args.max_template_date,
-        max_hits=config.data.predict.max_templates,
-        kalign_binary_path=args.kalign_binary_path,
-        release_dates_path=args.release_dates_path,
-        obsolete_pdbs_path=args.obsolete_pdbs_path
-    )
-
     use_small_bfd = args.preset == 'reduced_dbs'  # (args.bfd_database_path is None)
     if use_small_bfd:
         assert args.bfd_database_path is not None
@@ -357,8 +348,6 @@ def inference_monomer_model(args):
         assert args.bfd_database_path is not None
         assert args.uniref30_database_path is not None
 
-    data_processor = data_pipeline.DataPipeline(template_featurizer=template_featurizer,)
-
     output_dir_base = args.output_dir
     
     random_seed = args.data_random_seed
@@ -423,11 +412,31 @@ def inference_monomer_model(args):
                 use_small_bfd=use_small_bfd,
                 no_cpus=args.cpus,
             )
+            t = time.perf_counter()
             alignment_runner.run(fasta_path, local_alignment_dir)
+            print(f"Alignment data time: {time.perf_counter() - t}")
             
-    feature_dict = data_processor.process_fasta(fasta_path=fasta_path,
-                                            alignment_dir=local_alignment_dir)
-
+        features_output_path = os.path.join(local_alignment_dir, 'features.pkl')
+        if os.path.exists(features_output_path):
+            feature_dict = pickle.load(open(features_output_path, 'rb'))
+        
+        else:
+            template_featurizer = templates.TemplateHitFeaturizer(
+                mmcif_dir=args.template_mmcif_dir,
+                max_template_date=args.max_template_date,
+                max_hits=config.data.predict.max_templates,
+                kalign_binary_path=args.kalign_binary_path,
+                release_dates_path=args.release_dates_path,
+                obsolete_pdbs_path=args.obsolete_pdbs_path
+            )
+            
+            data_processor = data_pipeline.DataPipeline(template_featurizer=template_featurizer,)
+    
+            feature_dict = data_processor.process_fasta(fasta_path=fasta_path,
+                                                    alignment_dir=local_alignment_dir)
+        with open(features_output_path, 'wb') as f:
+            pickle.dump(feature_dict, f, protocol=4)
+        
     # Remove temporary FASTA file
     os.remove(fasta_path)