diff --git a/.buildkite/pipeline.yml b/.buildkite/pipeline.yml
new file mode 100644
index 000000000..165864871
--- /dev/null
+++ b/.buildkite/pipeline.yml
@@ -0,0 +1,33 @@
+env:
+  SECRET_CODECOV_TOKEN: ""
+
+steps:
+  - label: "Julia v1 -- CUDA"
+    plugins:
+      - JuliaCI/julia#v1:
+          version: "1"
+      - JuliaCI/julia-test#v1: ~
+      - JuliaCI/julia-coverage#v1:
+          dirs:
+            - src
+            - ext
+    agents:
+      queue: "juliagpu"
+      cuda: "*"
+    if: build.message !~ /\[skip tests\]/
+    timeout_in_minutes: 30
+
+  - label: "Julia LTS -- CUDA"
+    plugins:
+      - JuliaCI/julia#v1:
+          version: "1.10" # "lts" isn't valid
+      - JuliaCI/julia-test#v1: ~
+      - JuliaCI/julia-coverage#v1:
+          dirs:
+            - src
+            - ext
+    agents:
+      queue: "juliagpu"
+      cuda: "*"
+    if: build.message !~ /\[skip tests\]/
+    timeout_in_minutes: 30
diff --git a/test/cuda/cu_adapt.jl b/test/cuda/cu_adapt.jl
new file mode 100644
index 000000000..24360f637
--- /dev/null
+++ b/test/cuda/cu_adapt.jl
@@ -0,0 +1,30 @@
+using Test
+using MPSKit
+using TensorKit
+using CUDA, cuTENSOR, Adapt
+
+@testset "Adapting to CuArray" begin
+    # start from random operators
+    L = 4
+    T = ComplexF64
+
+    for V in (ℂ^2, U1Space(0 => 1, 1 => 1))
+        O₁ = rand(T, V^L, V^L)
+        O₂ = rand(T, space(O₁))
+        O₃ = rand(real(T), space(O₁))
+
+        # create MPO and convert it back to see if it is the same
+        mpo₁ = adapt(CuArray, FiniteMPO(O₁))
+        mpo₂ = adapt(CuArray, FiniteMPO(O₂))
+        mpo₃ = adapt(CuArray, FiniteMPO(O₃))
+
+        @test isfinite(mpo₁)
+        @test isfinite(typeof(mpo₁))
+        @test GeometryStyle(typeof(mpo₁)) == FiniteChainStyle()
+        @test GeometryStyle(mpo₁) == FiniteChainStyle()
+        @test OperatorStyle(typeof(mpo₁)) == MPOStyle()
+        @test TensorKit.storagetype(mpo₁) == CuVector{T, 1, CUDA.DeviceMemory}
+        @test TensorKit.storagetype(mpo₂) == CuVector{T, 1, CUDA.DeviceMemory}
+        @test TensorKit.storagetype(mpo₃) == CuVector{T, 1, CUDA.DeviceMemory}
+    end
+end
diff --git a/test/runtests.jl b/test/runtests.jl
index f08104e88..c1c525ade 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -12,4 +12,12 @@ init_code = quote
 end
 
 args = parse_args(ARGS)
+is_buildkite = get(ENV, "BUILDKITE", "false") == "true"
+if is_buildkite
+    empty!(testsuite)
+    gpu_testsuite = find_tests(joinpath(@__DIR__, "cuda"))
+    append!(testsuite, gpu_testsuite)
+else
+    filter!(!(startswith("cuda") ∘ first), testsuite)
+end
 runtests(MPSKit, args; testsuite, init_code)