diff --git a/.buildkite/pipeline.yml b/.buildkite/pipeline.yml
index 9a9014b649..ef3363f699 100644
--- a/.buildkite/pipeline.yml
+++ b/.buildkite/pipeline.yml
@@ -35,43 +35,6 @@ steps:
         if: build.message !~ /\[skip tests\]/
         timeout_in_minutes: 120
 
-      - label: ":julia: :linux: aarch64 - Julia v{{matrix.version}} -- {{matrix.group}}"
-        matrix:
-          setup:
-            version:
-              - "1.10"
-              - "1.11"
-            group:
-              - core
-              - neural_networks
-              - integration
-        plugins:
-          - JuliaCI/julia#v1:
-              version: "{{matrix.version}}"
-          - JuliaCI/julia-coverage#v1:
-              codecov: true
-              dirs:
-                - src
-                - ext
-                - lib/ReactantCore/src
-        commands: |
-          julia --project=. -e 'println("--- :julia: Instantiating project")
-              using Pkg
-              Pkg.develop([PackageSpec(path="lib/ReactantCore")])'
-
-          julia --project=. -e 'println("--- :julia: Run Tests")
-              using Pkg
-              Pkg.test(; coverage="user")'
-        agents:
-          queue: "juliaecosystem"
-          os: "linux"
-          sandbox_capable: "true"
-          arch: "aarch64"
-        env:
-          REACTANT_TEST_GROUP: "{{matrix.group}}"
-        if: build.message !~ /\[skip tests\]/
-        timeout_in_minutes: 120
-
   # - group: ":racehorse: Benchmarks"
   #   steps:
   #     - label: "CPU: Run Benchmarks"
diff --git a/.github/workflows/CI.yml b/.github/workflows/CI.yml
index f500982f15..bd7102079e 100644
--- a/.github/workflows/CI.yml
+++ b/.github/workflows/CI.yml
@@ -43,6 +43,7 @@ jobs:
           # - 'nightly'
         os:
           - ubuntu-20.04
+          - ubuntu-24.04-arm
           - macOS-latest
         test_group:
           - core
@@ -89,9 +90,10 @@ jobs:
           #   version: '1.10'
           #   test_group: integration
         exclude:
-          # these are run on Buildkite
-          - os: ubuntu-20.04
+          - os: ubuntu-20.04 # this is x86_64, exclude foreign architecture
             arch: aarch64
+          - os: ubuntu-24.04-arm # this is aarch64, exclude foreign architecture
+            arch: x64
     steps:
       - uses: actions/checkout@v4
       - uses: julia-actions/setup-julia@v2
diff --git a/test/integration/cuda.jl b/test/integration/cuda.jl
index 4ec52ab869..8cb3475bbc 100644
--- a/test/integration/cuda.jl
+++ b/test/integration/cuda.jl
@@ -16,6 +16,12 @@ function square!(x, y)
     return nothing
 end
 
+# GPUCompiler throws "Not implemented" errors on aarch64 before
+# <https://github.com/JuliaLang/julia/pull/57077> for some tests.
+const skip_tests =
+    Base.BinaryPlatforms.arch(Base.BinaryPlatforms.HostPlatform()) == "aarch" &&
+    VERSION <= v"1.11.2"
+
 @static if !Sys.isapple()
     @testset "Square Kernel" begin
         oA = collect(1:1:64)
@@ -26,7 +32,9 @@ end
             @test all(Array(A) .≈ (oA .* oA .* 100))
             @test all(Array(B) .≈ (oA .* 100))
         else
-            @code_hlo optimize = :before_kernel square!(A, B)
+            @static if !skip_tests
+                @code_hlo optimize = :before_kernel square!(A, B)
+            end
         end
     end
 end
@@ -53,7 +61,9 @@ end
             @test all(Array(A) .≈ oA .* sin.(oA .* 100))
             @test all(Array(B) .≈ (oA .* 100))
         else
-            @code_hlo optimize = :before_kernel sin!(A, B)
+            @static if !skip_tests
+                @code_hlo optimize = :before_kernel sin!(A, B)
+            end
         end
     end
 end
@@ -79,7 +89,9 @@ end
             @jit smul!(A)
             @test all(Array(A) .≈ oA .* 15)
         else
-            @code_hlo optimize = :before_kernel smul!(A)
+            @static if !skip_tests
+                @code_hlo optimize = :before_kernel smul!(A)
+            end
         end
     end
 end
@@ -104,7 +116,9 @@ tuplef2(a) = @cuda threads = 1 tuplef2!((5, a))
             @jit tuplef(A)
             @test all(Array(A) .≈ 3)
         else
-            @code_hlo optimize = :before_kernel tuplef(A)
+            @static if !skip_tests
+                @code_hlo optimize = :before_kernel tuplef(A)
+            end
         end
 
         A = ConcreteRArray(fill(1))
@@ -120,7 +134,9 @@ tuplef2(a) = @cuda threads = 1 tuplef2!((5, a))
         @jit tuplef2(A)
         @test all(Array(A) .≈ 5)
     else
-        @code_hlo optimize = :before_kernel tuplef2(A)
+        @static if !skip_tests
+            @code_hlo optimize = :before_kernel tuplef2(A)
+        end
     end
 end
 
diff --git a/test/runtests.jl b/test/runtests.jl
index 7d188fe3dd..be17750042 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -60,7 +60,6 @@ const REACTANT_TEST_GROUP = lowercase(get(ENV, "REACTANT_TEST_GROUP", "all"))
     end
 
     if REACTANT_TEST_GROUP == "all" || REACTANT_TEST_GROUP == "integration"
-        # Temporarily disabled as minutia are debugged
         @safetestset "CUDA" include("integration/cuda.jl")
         @safetestset "Linear Algebra" include("integration/linear_algebra.jl")
         @safetestset "OffsetArrays" include("integration/offsetarrays.jl")