diff --git a/docs/introduction/img/level-1-modelserving.png b/docs/introduction/img/level-1-modelserving.png
new file mode 100644
index 00000000..d4995ec6
Binary files /dev/null and b/docs/introduction/img/level-1-modelserving.png differ
diff --git a/docs/introduction/img/level-2.png b/docs/introduction/img/level-2.png
index 8d400751..901225a2 100644
Binary files a/docs/introduction/img/level-2.png and b/docs/introduction/img/level-2.png differ
diff --git a/docs/introduction/levels.md b/docs/introduction/levels.md
index ae8ce7fb..42609ea7 100644
--- a/docs/introduction/levels.md
+++ b/docs/introduction/levels.md
@@ -4,7 +4,7 @@ description: "Levels of MLOps"
 sidebar_position: 2
 date: 2021-12-03
 lastmod: 2022-03-05
-contributors: ["Jongseob Jeon"]
+contributors: ["Jongseob Jeon", "Chanmin Cho"]
 
 ---
 
@@ -81,6 +81,13 @@ Real World에서 데이터는 Data Shift라는 데이터의 분포가 계속해
 
 정리하자면 CT를 위해서는 Auto Retraining과 Auto Deploy 두 가지 기능이 필요합니다. 둘은 서로의 단점을 보완해 계속해서 모델의 성능을 유지할 수 있게 합니다.
 
+### Model Serving
+
+![level-1-modelserving](./img/level-1-modelserving.png)
+
+프로덕션 환경에서의 머신러닝 파이프라인은 새로운 데이터에 기반한 최신 모델을 예측 서비스에 지속적으로 배포합니다. 이 과정에서, 훈련되고 검증된 모델을 온라인 예측 서비스에 자동적으로 배포하는 작업이 포함됩니다.
+
+
 ## 2단계: CI/CD 파이프라인의 자동화
 
 ![level-2](./img/level-2.png)
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/introduction/img/level-1-modelserving.png b/i18n/en/docusaurus-plugin-content-docs/current/introduction/img/level-1-modelserving.png
new file mode 100644
index 00000000..d4995ec6
Binary files /dev/null and b/i18n/en/docusaurus-plugin-content-docs/current/introduction/img/level-1-modelserving.png differ
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/introduction/img/level-2.png b/i18n/en/docusaurus-plugin-content-docs/current/introduction/img/level-2.png
index 8d400751..901225a2 100644
Binary files a/i18n/en/docusaurus-plugin-content-docs/current/introduction/img/level-2.png and b/i18n/en/docusaurus-plugin-content-docs/current/introduction/img/level-2.png differ
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/introduction/levels.md b/i18n/en/docusaurus-plugin-content-docs/current/introduction/levels.md
index 403e1cf2..d6a8f5c9 100644
--- a/i18n/en/docusaurus-plugin-content-docs/current/introduction/levels.md
+++ b/i18n/en/docusaurus-plugin-content-docs/current/introduction/levels.md
@@ -4,7 +4,7 @@ description: "Levels of MLOps"
 sidebar_position: 2
 date: 2021-12-03
 lastmod: 2022-03-05
-contributors: ["Jongseob Jeon"]
+contributors: ["Jongseob Jeon", "Chanmin Cho"]
 
 ---
 
@@ -77,6 +77,12 @@ There is a simple solution to address this blind spot. It involves checking whet
 To summarize, for Continuous Training (CT), both Auto Retrain and Auto Deploy are necessary. They complement each other's weaknesses and enable the model's performance to be maintained continuously.
 
 
+### Model Serving
+
+![level-1-modelserving](./img/level-1-modelserving.png)
+
+Machine learning pipelines in production continuously deploy the latest models based on new data to your prediction service. This process involves automatically deploying trained and validated models to online prediction services.
+
 
 ## Level 2: Automating the CI/CD Pipeline