diff --git a/Dockerfile b/Dockerfile
new file mode 100644
index 0000000..b88cad5
--- /dev/null
+++ b/Dockerfile
@@ -0,0 +1,24 @@
+# Use an official Python 3.11 runtime as a parent image
+FROM python:3.11-slim
+
+# Set the working directory
+WORKDIR /app
+
+# Copy the current directory contents into the container
+COPY . /app
+
+# This tells Python to look in /app for the 'recml' package
+ENV PYTHONPATH="${PYTHONPATH}:/app"
+
+# Install system tools if needed (e.g., git)
+RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*
+
+# Install dependencies
+RUN pip install --upgrade pip
+RUN pip install -r requirements.txt
+
+# Force install the specific protobuf version
+RUN pip install "protobuf>=6.31.1" --no-deps
+
+# Default command to run the training script
+CMD ["python", "recml/examples/dlrm_experiment_test.py"]
diff --git a/training.md b/training.md
index cb82670..bf36dbc 100644
--- a/training.md
+++ b/training.md
@@ -6,7 +6,7 @@ This guide explains how to set up the environment and train the HSTU/DLRM models
 
 If you are developing on a TPU VM directly, use a virtual environment to avoid conflicts with the system-level Python packages.
 
-#### 1. Prerequisites
+### 1. Prerequisites
 Ensure you have **Python 3.11+** installed.
 ```bash
 python3 --version
@@ -41,7 +41,7 @@ python dlrm_experiment_test.py
 
 If you prefer not to manage a virtual environment or want to deploy this as a container, you can build a Docker image.
 
-## 1. Build the Image
+### 1. Create a Dockerfile
 Create a file named `Dockerfile` in the root of the repository:
 
 ```dockerfile
@@ -54,6 +54,9 @@ WORKDIR /app
 # Copy the current directory contents into the container
 COPY . /app
 
+# This tells Python to look in /app for the 'recml' package
+ENV PYTHONPATH="${PYTHONPATH}:/app"
+
 # Install system tools if needed (e.g., git)
 RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*
 
@@ -68,4 +71,28 @@ RUN pip install "protobuf>=6.31.1" --no-deps
 CMD ["python", "recml/examples/dlrm_experiment_test.py"]
 ```
 
-You can use this dockerfile to run the DLRM model experiment from this repo in your own environment. 
\ No newline at end of file
+You can use this dockerfile to run the DLRM model experiment from this repo in your own environment. 
+
+### 2. Build the Image
+
+Run this command from the root of the repository. It reads the `Dockerfile`, installs all dependencies, and creates a ready-to-run image.
+
+```bash
+docker build -t recml-training .
+```
+
+### 3. Run the Image
+
+```bash
+docker run --rm --privileged \
+  --net=host \
+  --ipc=host \
+  --name recml-experiment \
+  recml-training
+```
+
+### What is happening here?
+* **`--rm`**: Automatically deletes the container after the script finishes to keep your disk clean.
+* **`--privileged`**: Grants the container direct access to the host's hardware devices, which is required to see the physical TPU chips.
+* **`--net=host`**: Removes the container's network isolation, allowing the script to connect to the TPU runtime listening on local ports (e.g., 8353).
+* **`--ipc=host`**: Allows the container to use the host's Shared Memory (IPC), which is critical for high-speed data transfer between the CPU and TPU.
\ No newline at end of file