diff --git a/arm_compute/dynamic_fusion/sketch/MemoryDescriptor.h b/arm_compute/dynamic_fusion/sketch/MemoryDescriptor.h
new file mode 100644
index 0000000..deedf62
--- /dev/null
+++ b/arm_compute/dynamic_fusion/sketch/MemoryDescriptor.h
@@ -0,0 +1,76 @@
+/*
+ * Copyright (c) 2022 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_MEMORYDESCRIPTOR
+#define ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_MEMORYDESCRIPTOR
+
+#include "arm_compute/core/ITensorInfo.h"
+
+namespace arm_compute
+{
+namespace experimental
+{
+namespace dynamic_fusion
+{
+/** Type of memory used by a workload tensor */
+enum class MemoryType
+{
+    User      = 0, /**< Memory coming directly from users, e.g. for argument tensors */
+    Auxiliary = 1, /**< Additional memory required by the workload tensor, e.g. for temporary tensors */
+};
+
+/** Memory information for tensors with @ref MemoryType::Auxiliary.
+ * This informs how much additional memory is required for auxiliary tensors
+ */
+struct AuxMemoryInfo
+{
+    AuxMemoryInfo() = default;
+
+    AuxMemoryInfo(size_t size, size_t alignment = 0) noexcept
+        : size(size),
+          alignment(alignment)
+    {
+    }
+
+    friend bool operator==(const AuxMemoryInfo &info0, const AuxMemoryInfo &info1)
+    {
+        return info0.size == info1.size && info0.alignment == info1.alignment;
+    }
+    size_t size{ 0 };      /**< Total memory size in bytes */
+    size_t alignment{ 0 }; /**< Memory alignment in bytes */
+};
+
+/** Descriptor of a workload tensor memory */
+struct MemoryDescriptor
+{
+    MemoryType    memory_type{};     /**< Memory Type*/
+    AuxMemoryInfo aux_memory_info{}; /**< Auxiliary Tensor Memory Information */
+};
+
+/** A map from @ref ITensorInfo to their corresponding @ref MemoryDescriptor */
+using MemoryDescriptorMap = std::map<ITensorInfo::Id, MemoryDescriptor>;
+
+} // namespace dynamic_fusion
+} // namespace experimental
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_MEMORYDESCRIPTOR */
diff --git a/arm_compute/dynamic_fusion/sketch/OperatorAttributes.h b/arm_compute/dynamic_fusion/sketch/OperatorAttributes.h
new file mode 100644
index 0000000..22c6772
--- /dev/null
+++ b/arm_compute/dynamic_fusion/sketch/OperatorAttributes.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright (c) 2022 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_OPERATORATTRIBUTES
+#define ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_OPERATORATTRIBUTES
+
+#include "arm_compute/core/Size2D.h"
+#include "arm_compute/core/Types.h"
+#include <cstdint>
+
+namespace arm_compute
+{
+namespace experimental
+{
+namespace dynamic_fusion
+{
+/** Attributes are backend-agnostic parameters (in addition to the input/output tensors) of an operator.
+ */
+
+/** Conv2d attributes */
+class Conv2dAttributes
+{
+public:
+    /** Set padding */
+    Conv2dAttributes &pad(const Padding2D &pad);
+    /** Get padding */
+    Padding2D pad() const;
+    /** Set stride */
+    Conv2dAttributes &stride(const Size2D &stride);
+    /** Get stride */
+    Size2D stride() const;
+    /** Set dilation */
+    Conv2dAttributes &dilation(const Size2D &dilation);
+    /** Get dilation */
+    Size2D dilation() const;
+
+private:
+    Padding2D _pad{};              /**< Padding */
+    Size2D    _stride{ 1U, 1U };   /**< Stride */
+    Size2D    _dilation{ 1U, 1U }; /**< Dilation */
+};
+} // namespace dynamic_fusion
+} // namespace experimental
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_OPERATORATTRIBUTES */
diff --git a/arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadContext.h b/arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadContext.h
new file mode 100644
index 0000000..1ee3c7e
--- /dev/null
+++ b/arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadContext.h
@@ -0,0 +1,87 @@
+/*
+ * Copyright (c) 2022 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADCONTEXT
+#define ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADCONTEXT
+
+#include "arm_compute/core/GPUTarget.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+/** Forward declaration */
+class CLCompileContext;
+namespace experimental
+{
+namespace dynamic_fusion
+{
+/** Gpu Information such as the Gpu target (for example, G76) */
+using GpuTarget = ::arm_compute::GPUTarget;
+
+/** Gpu Language */
+enum class GpuLanguage
+{
+    OpenCL,
+    Unknown
+};
+/** Provide context necessary for the creation and configuration of a workload
+ * e.g. gpu targets and capabilities, cl::Device for querying OpenCl extensions. Both can affect how a kernel is generated
+ *
+ * This context is shared between different operators within a sketch, and has to stay valid for the entire workload creation session.
+ * This context may also be shared between different sketches.
+ *
+ * This class only contains information for workload creation, but not for runtime (e.g. cl::Queue for enqueueing the kernels)
+ */
+class GpuWorkloadContext
+{
+public:
+    /** Constructor */
+    GpuWorkloadContext(CLCompileContext *cl_compile_context);
+    /** Allow instances of this class to be copy constructed */
+    GpuWorkloadContext(const GpuWorkloadContext &config) = default;
+    /** Allow instances of this class to be copied */
+    GpuWorkloadContext &operator=(const GpuWorkloadContext &config) = default;
+    /** Allow instances of this class to be move constructed */
+    GpuWorkloadContext(GpuWorkloadContext &&config) = default;
+    /** Allow instances of this class to be moved */
+    GpuWorkloadContext &operator=(GpuWorkloadContext &&config) = default;
+    /** Get @ref GpuLanguage of the context */
+    GpuLanguage gpu_language() const;
+    /** Get @ref GpuTarget of the context */
+    GpuTarget gpu_target() const;
+    /** Get @ref CLCompileContext
+     * If the gpu language is not OpenCL, then return nullptr
+     */
+    const CLCompileContext *cl_compile_context() const;
+
+private:
+    GpuLanguage       _gpu_language{ GpuLanguage::Unknown };
+    CLCompileContext *_cl_compile_ctx{ nullptr };
+};
+
+} // namespace dynamic_fusion
+} // namespace experimental
+} // namespace arm_compute
+
+#endif /* ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADCONTEXT */
diff --git a/arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadSketch.h b/arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadSketch.h
new file mode 100644
index 0000000..afbe2b8
--- /dev/null
+++ b/arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadSketch.h
@@ -0,0 +1,100 @@
+/*
+ * Copyright (c) 2022 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADSKETCH
+#define ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADSKETCH
+
+#include "arm_compute/core/TensorInfo.h"
+#include "arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadContext.h"
+
+#include <memory>
+
+namespace arm_compute
+{
+namespace experimental
+{
+namespace dynamic_fusion
+{
+/** A descriptor of a workload of operators
+ *
+ * A "workload" is a basic unit of computation to schedule and perform. It contains one or more operators that can be "fused" together.
+ * Note that a workload may still contain multiple kernels.
+ */
+class GpuWorkloadSketch
+{
+public:
+    /** Global context used for the creation of a workload */
+    using Context = GpuWorkloadContext;
+    /** Internal opaque implementation */
+    class Implementation;
+
+public:
+    /** Constructor
+     *
+     * @param[in] context Gpu context for the creation of a workload
+     */
+    explicit GpuWorkloadSketch(GpuWorkloadContext *context);
+    /** Destructor */
+    ~GpuWorkloadSketch();
+    /** Get the implementation */
+    Implementation &implementation();
+    /** Get the implementation */
+    const Implementation &implementation() const;
+    /** Get the gpu workload context of this sketch */
+    const GpuWorkloadContext *gpu_context() const;
+    /** Create a @ref TensorInfo associated with the workload sketch.
+     *
+     * @return TensorInfo   Newly created tensor info
+     */
+    template <typename... Args>
+    TensorInfo create_tensor_info(Args &&... args)
+    {
+        auto tensor_info = TensorInfo(std::forward<Args>(args)...);
+        tensor_info.set_id(allocate_new_tensor_id());
+        return tensor_info;
+    }
+    /** Create a @ref TensorInfo associated with the workload sketch by copying from an existing tensor info
+     * @note The newly copied tensor will have a different identity within the workload than the one copied from
+     *       To copy the identity of @p tensor_info as well, use @ref TensorInfo 's copy constructors instead
+     *
+     * @param[in] tensor_info @ref ITensorInfo to copy from
+     *
+     * @return TensorInfo   Newly created tensor info
+     */
+    TensorInfo create_tensor_info(const ITensorInfo &tensor_info);
+    /** Create a default @ref TensorInfo associated with the workload sketch
+     * It is usually used by a destination tensor whose @ref ITensorInfo is to be inferred automatically
+     *
+     * @return TensorInfo   Newly created tensor info
+     */
+    TensorInfo create_tensor_info();
+
+private:
+    ITensorInfo::Id                 allocate_new_tensor_id();
+    std::unique_ptr<Implementation> _impl; /**< Internal opaque implementation*/
+};
+
+} // namespace dynamic_fusion
+} // namespace experimental
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_GPUWORKLOADSKETCH */
diff --git a/arm_compute/dynamic_fusion/sketch/gpu/operators/GpuConv2d.h b/arm_compute/dynamic_fusion/sketch/gpu/operators/GpuConv2d.h
new file mode 100644
index 0000000..fe9108d
--- /dev/null
+++ b/arm_compute/dynamic_fusion/sketch/gpu/operators/GpuConv2d.h
@@ -0,0 +1,84 @@
+/*
+ * Copyright (c) 2022 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_OPERATORS_GPUCONV2D
+#define ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_OPERATORS_GPUCONV2D
+
+#include "arm_compute/core/Error.h"
+#include "arm_compute/dynamic_fusion/sketch/OperatorAttributes.h"
+
+namespace arm_compute
+{
+namespace experimental
+{
+namespace dynamic_fusion
+{
+/** Forward declaration */
+class GpuWorkloadSketch;
+
+/** Operator interface. */
+class GpuConv2d final
+{
+public:
+    /** Attributes are a set of backend-agnostic parameters that define what an operator does */
+    using Attributes = Conv2dAttributes;
+    /** Create an operator and fuse it into the workload sketch.
+     *    @note If @ref validate_op() fails, the creation also fails and may throw an error.
+     *    @note If @ref validate_op() fails, @p sketch remains unchanged and valid.
+     *
+     * Valid data type configurations:
+     * |src            |wei            |bia            |dst            |
+     * |:--------------|:--------------|:--------------|:--------------|
+     * |F16            |F16            |F16            |F16            |
+     * |F32            |F32            |F32            |F32            |
+     *
+     * Valid data layouts:
+     * - NHWC
+     *
+     * @param[in,out] sketch     Workload sketch into which the operator will be fused
+     * @param[in]     src        Source tensor
+     * @param[in]     wei        Weight tensor
+     * @param[in]     bia        (Optional) Bias tensor
+     * @param[out]    dst        Destination tensor
+     * @param[in]     attributes Operator attributes
+     */
+    static void create_op(GpuWorkloadSketch &sketch,
+                          ITensorInfo       *src,
+                          ITensorInfo       *wei,
+                          ITensorInfo       *bia,
+                          ITensorInfo       *dst,
+                          const Attributes &attributes);
+    /** Validate the operator and check if it can be fused into the workload sketch.
+     * Similar to @ref GpuConv2d::create_op()
+     */
+    static Status validate_op(const GpuWorkloadSketch &sketch,
+                              const ITensorInfo       *src,
+                              const ITensorInfo       *wei,
+                              const ITensorInfo       *bia,
+                              const ITensorInfo       *dst,
+                              const Attributes        &attributes);
+};
+} // namespace dynamic_fusion
+} // namespace experimental
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_DYNAMIC_FUSION_SKETCH_GPU_OPERATORS_GPUCONV2D */