About

Hi, I’m Alex - an AI/ML Software Engineer with end-to-end, full-stack experience in porting, optimizing, and deploying AI/ML models on RISC-V platforms, spanning the full pipeline from frontend models (PyTorch/ONNX) through MLIR-based compiler lowering and code generation, runtime and driver integration, and deployment on RISC-V hardware.

Hands-on experience in porting, validating, and optimizing diverse AI workloads on RISC-V platforms, especially for LLMs.

Experienced with mainstream AI frameworks such as PyTorch and ONNX, including custom operator registration and integration, as well as replacing CUDA/C++ kernels with RISC-V implementations.

Strong background in the MLIR ecosystem, including Torch-MLIR for lowering PyTorch models into MLIR, and IREE as an MLIR-based AI compiler and runtime, with experience in RISC-V and GPU code generation, custom kernel development, and integration of custom runtimes for SiFive accelerators.

Solid understanding of PyTorch JIT compilation, including TorchDynamo mechanisms, graph capture, and supporting torch.compile with a RISC-V backend.

Familiar with modern LLM inference frameworks, including vLLM and llama.cpp, with hands-on experience evaluating and integrating them into end-to-end LLM workflows.

Hands-on experience with distributed computing and multi-device LLM execution, including model sharding via Tensor Parallelism (TP).

You can find my LinkedIn profile here: https://www.linkedin.com/in/jyx1118/