diff --git a/.jules/bolt.md b/.jules/bolt.md
new file mode 100644
index 0000000..e8b5081
--- /dev/null
+++ b/.jules/bolt.md
@@ -0,0 +1,3 @@
+## 2024-05-24 - [Cache Unnecessary Python Loop Variables]
+**Learning:** In highly recursive or looping NLP functions such as line-by-line regex scanning, rebuilding configurations or alias lookups dynamically creates enormous overhead in Python. The openmed package's `_alias_lookups` recreated language alias dicts on every single line, leading to severe slowdowns.
+**Action:** Use `@functools.lru_cache(maxsize=None)` on deterministic configuration generators within loops to cut down redundant operations. Ensure inputs are hashable and small in variance (like language codes).
diff --git a/openmed/openmed/clinical/sections/detect.py b/openmed/openmed/clinical/sections/detect.py
index 8716746..1d9b637 100644
--- a/openmed/openmed/clinical/sections/detect.py
+++ b/openmed/openmed/clinical/sections/detect.py
@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import functools
 from collections.abc import Iterable, Mapping
 from dataclasses import dataclass
 from typing import Any
@@ -244,6 +245,9 @@ def _is_underline(text: str) -> bool:
     return len(stripped) >= 3 and set(stripped) <= _UNDERLINE_CHARS
 
 
+# Bolt ⚡: Cache alias lookups as they are static per language to prevent
+# redundant tuple/dict allocations on every line during clinical section detection.
+@functools.lru_cache(maxsize=None)
 def _alias_lookups(language: str | None) -> tuple[tuple[str, Mapping[str, str]], ...]:
     languages = (
         tuple(dict.fromkeys((get_section_lexicon(language).language, "en")))
diff --git a/openmed/uv.lock b/openmed/uv.lock
index 78f4815..430a3da 100644
--- a/openmed/uv.lock
+++ b/openmed/uv.lock
@@ -4439,6 +4439,7 @@ wheels = [
 name = "openmed"
 source = { editable = "." }
 dependencies = [
+    { name = "defusedxml" },
     { name = "faker" },
     { name = "pysbd" },
     { name = "pyyaml" },
@@ -4622,6 +4623,7 @@ requires-dist = [
     { name = "coremltools", marker = "extra == 'coreml'", specifier = ">=8.0" },
     { name = "dask", extras = ["dataframe"], marker = "extra == 'dask'", specifier = ">=2024.8" },
     { name = "dask", extras = ["dataframe"], marker = "extra == 'dev'", specifier = ">=2024.8" },
+    { name = "defusedxml", specifier = ">=0.7.1" },
     { name = "duckdb", marker = "extra == 'dev'", specifier = ">=1.0,<2" },
     { name = "duckdb", marker = "extra == 'duckdb'", specifier = ">=1.0,<2" },
     { name = "easyocr", marker = "extra == 'multimodal'", specifier = ">=1.7" },