diff --git a/docs/architecture.md b/docs/architecture.md
index 08436d7..76480c3 100644
--- a/docs/architecture.md
+++ b/docs/architecture.md
@@ -15,7 +15,7 @@ This page contains a high-level overview of SlateDB's architecture, read/write p
 
 The following diagram shows the architecture of SlateDB:
 
-![Example banner](/img/architecture.png)
+![Architecture diagram](/img/architecture.png)
 
 At a high level, SlateDB consists of the following components:
 
@@ -29,6 +29,8 @@ At a high level, SlateDB consists of the following components:
 
 ## Writes
 
+![Write diagram](/img/architecture-write.png)
+
 SlateDB's write path is as follows:
 
 1. A `put` call is made on the client.
@@ -39,6 +41,8 @@ SlateDB's write path is as follows:
 
 ## Reads
 
+![Read diagram](/img/architecture-read.png)
+
 SlateDB's read path is as follows:
 
 1. A `get` call is made on the client.
@@ -69,6 +73,8 @@ SlateDB's manifest file contains the current state of the database, including:
 
 ## Compaction
 
+![Compaction diagram](/img/architecture-compaction.png)
+
 L0 SSTs are written to the `compacted` directory in the object store when `l0_sst_size_bytes` is exceeded. SlateDB's compactor is responsible for merging SSTs from L0 into lower levels (L1, L2, and so on). These lower levels are referred to as _sorted runs_ in SlateDB. Each SST in a sorted run contains a distinct subset of the keyspace.
 
 SlateDB's compactor has the following components:
diff --git a/static/img/architecture-compaction.png b/static/img/architecture-compaction.png
new file mode 100644
index 0000000..c8cd59b
Binary files /dev/null and b/static/img/architecture-compaction.png differ
diff --git a/static/img/architecture-read.png b/static/img/architecture-read.png
new file mode 100644
index 0000000..b5294a5
Binary files /dev/null and b/static/img/architecture-read.png differ
diff --git a/static/img/architecture-write.png b/static/img/architecture-write.png
new file mode 100644
index 0000000..b353cd4
Binary files /dev/null and b/static/img/architecture-write.png differ