发现心仪岗位

Role Overview

Design compiler technologies that translate AI models into optimized code for the RISC-V AI SoC. You will focus on IR design, code generation, autotuning, and ensuring clean integration with upstream toolchains (LLVM, MLIR, TVM).

Key Responsibilities

1. Extend MLIR/LLVM to support vector and matrix instruction sets.

2. Implement optimization and lowering passes for AI operators.

3. Build compiler interfaces compatible with ecosystem conventions.

4. Integrate autotuning and performance modeling.

5. Collaborate with runtime and kernel teams to align code generation strategies.

Qualifications

1. 4+ years in compiler backend or performance optimization.

2. Strong C++/Python proficiency.

3. Familiarity with MLIR, LLVM, TVM, or XLA.

Preferred Attributes

1. Experience integrating custom instructions or intrinsics 

2. Knowledge of AI model graph-to-code lowering workflows.

3. Familiarity with autotuning systems like AutoTVM or MLIR-based tuners.

4. Understanding of mixed-precision compute and performance trade-offs.

Impact: You enable portable, high-performance compilation pipelines that power ecosystem integration.

Role Overview

You will develop, optimize, and demonstrate end-to-end AI applications and reference designs on the proprietary RISC-V AI SoC and SDK. This includes benchmarking, demos, and workload enablement to showcase platform capabilities and provide ready-to-use examples for developers and customers.

Key Responsibilities

1. Develop and port AI workloads (CNN, Transformer, multimodal) using PyTorch, TensorFlow, and ONNX on the SDK. 

2. Create reference designs combining compiler, runtime, and kernel components. 

3. Develop SDK demos, tutorials, and reproducible benchmarking pipelines. 

4. Optimize models for performance, accuracy, and power efficiency. 

5. Collaborate with compiler, runtime, and framework teams to debug and tune workloads. 

6. Deliver performance analysis reports and reference application repositories. 

7. Collect usability feedback from internal and external developers. 

Qualifications

1. 3–7 years of experience in AI model development, optimization, or deployment. 

2. Strong hands-on with PyTorch, TensorFlow, or ONNX Runtime. 

3. Experience running AI workloads on GPU, NPU, or heterogeneous compute systems. 

4. Strong Python skills; C++ proficiency preferred. 

Preferred Attributes

1. Familiarity with quantization, mixed precision, and compiler-based model optimization. 

2. Experience creating AI SDK demos, sample apps, or developer documentation. 

3. Understanding of CNN, transformer, and attention-based models.