diff --git a/docs/function_abi.md b/docs/function_abi.md
index bfd2002..ec9e1d1 100644
--- a/docs/function_abi.md
+++ b/docs/function_abi.md
@@ -6,7 +6,7 @@
 this requires additional metadata on top of the raw characteristics of a
 function. Where possible, this is done by attaching attributes to a function.
 
-*   `abi` : string indiciating the abi/calling convention in use
+*   `abi` : string indicating the abi/calling convention in use
 *   `abiv` : numeric version of the `abi`
 
 Each abi can require additional attributes as needed.
@@ -132,7 +132,7 @@
 compatible memory layout is on the host-language binding. However, often it is
 the most efficient to schedule this for execution on a device. In the future, it
 is anticipated that there will be a built-in pathway for scheduling such a
-conversion (which would allow pipelinining and offload of buffer conversions).
+conversion (which would allow pipelining and offload of buffer conversions).
 
 ##### Deferred result allocation
 
diff --git a/docs/simple_ir_walkthrough.md b/docs/simple_ir_walkthrough.md
index c728234..331d5db 100644
--- a/docs/simple_ir_walkthrough.md
+++ b/docs/simple_ir_walkthrough.md
@@ -201,7 +201,7 @@
 that can be entirely elided, and reorder dispatches so that they can more easily
 be grouped based on required barriers. Thanks to MLIR's built-in folding logic
 we can also do some IR optimizations such as converting the generic dispatch to
-a `iree_ll_seq.static_dispatch`, as we know the workkload size at compile-time.
+a `iree_ll_seq.static_dispatch`, as we know the workload size at compile-time.
 
 As part of this we also propagate any static information we can determine, such
 as the workload, into the executables. This is to help aid backends in lowering