Skip to main content
mshr-h.com

Curriculum Vitae

Masahiro Hiramori (平森 将裕)

E-mail: contact@mshr-h.com
Location: Greater Tokyo Area (Kanagawa), Japan
GitHub: mshr-h
LinkedIn: masahiro-hiramori

📄 Download CV as PDF

Languages: Japanese (Native), English (Full professional proficiency)
Nationality: Japanese

Summary

AI Compiler / Edge Inference engineer with 8+ years of experience spanning AI compilers, runtime systems, and FPGA/ASIC-based acceleration for industrial and manufacturing workloads. Currently Head Engineer in Mitsubishi Electric’s Open-Source & InnerSource Program Office (OSPO), leading company-wide technical direction for open-source adoption across AI compilers, LLM deployment, and edge inference.

PyTorch Ambassador and Apache TVM Committer, with a proven record of bridging PyTorch ↔ TVM via modern export/import workflows and maintaining production-grade CI for compiler toolchains. Hands-on experience deploying vision models and domain-specific LLMs on edge devices including NVIDIA Jetson Orin Nano, Rockchip RK3588, Hailo-8, and Renesas RZ/V2H, and building runtime stacks for FPGA-based accelerators.

Skills

Languages

  • C, C++, Python

AI Frameworks & Compilers

  • PyTorch, Apache TVM (Relay / Relax / MetaSchedule), ONNX / ONNX-MLIR, Model optimization and compilation pipelines for edge devices, Integration of torch.export.ExportedProgram with Apache TVM

Hardware & Edge Inference

  • AI accelerators: Hailo-8, Edge SoCs: NVIDIA Jetson Orin Series, Rockchip RK3588, Renesas RZ/V2H, Performance tuning and deployment of object detection, anomaly detection, and domain-specific LLMs on edge devices

DevOps & Tooling

  • CI/CD (e.g., GitHub Actions), test automation, Reproducible ML workflows, containerization, continuous integration for compiler/tooling codebases

Open Source & Community

  • Large-scale OSS contribution workflows, code review, and maintainer responsibilities, Program committee and community leadership (PyTorch Conference, PyTorch Community Awards)

Experience

PyTorch Ambassador, PyTorch Community — Oct 2025 – Present

  • Serve as a PyTorch Ambassador representing industrial and edge AI compiler use cases.
  • Promote best practices for PyTorch deployment, compiler integration, and community contributions through talks, articles, and mentoring.

Apache TVM Committer, Apache Software Foundation — Sep 2024 – Present

  • Committer and core contributor to Apache TVM, focusing on PyTorch frontend integration, Relax, and MetaSchedule for inference workloads.
  • Maintain CI pipelines and review contributions related to PyTorch integration and model import/export workflows.

Head Engineer, Open-Source & InnerSource Program Office, Mitsubishi Electric Corporation, Japan — Apr 2025 – Present

  • Lead company-wide strategy for adopting open-source AI frameworks and tools, defining technical direction and governance for AI compilers, LLMs, and edge inference.
  • Drive deployment of a manufacturing domain-specific LLM to edge devices (NVIDIA Jetson Orin Nano, Rockchip RK3588), coordinating cross-functional teams to validate feasibility and align with production requirements.
  • Represent the company in external OSS communities (PyTorch, Apache TVM), aligning internal R&D roadmap with upstream projects and contributing back improvements.
  • Provide technical oversight and mentoring on AI/ML infrastructure, edge deployment, and OSS contribution workflows.

Researcher, Information Technology R&D Center, Mitsubishi Electric Corporation, Japan — Apr 2017 – Mar 2025

  • Developed a C++ user-mode driver and runtime stack for a proprietary FPGA-based neural network accelerator, achieving 100+ fps object detection on Xilinx UltraScale+ MPSoC in real-time scenarios.
  • Implemented image anomaly detection pipelines on edge AI chips such as Hailo-8 and Renesas RZ/V2H for visual inspection in manufacturing use cases, optimizing models and runtime for real-time performance on resource-constrained devices.
  • Implemented deployment of a manufacturing domain-specific LLM on edge devices (NVIDIA Jetson Orin Nano, Rockchip RK3588), including model optimization and runtime integration for on-device inference.
  • Prototyped end-to-end software systems for edge AI, including model conversion, compilation, and deployment pipelines using TVM and related tooling.
  • Contributed to open-source projects such as Apache TVM, Hummingbird, and ONNX-MLIR, focusing on PyTorch frontend development, CI/CD infrastructure, and integration with industrial edge workloads.

Open-Source & Community Activities

Apache TVM (Committer)

PyTorch

  • Contributor and committer to PyTorch
  • Implemented and maintained integration paths between PyTorch and Apache TVM to offload models for optimized inference on edge devices.

Microsoft Hummingbird

  • Contributor and committer to Microsoft Hummingbird
  • Integrated Apache TVM as a backend for classical ML model conversion.
  • Extended CI pipelines to validate conversion and execution of models targeting TVM-based runtimes.

Selected Projects

vscode-verilog-hdl-support, Maintainer — github.com/mshr-h/vscode-verilog-hdl-support

  • Maintainer of one of the most widely used Verilog/SystemVerilog HDL extensions for Visual Studio Code. Provides syntax highlighting, basic language support, and integration with typical hardware design workflows used in FPGA/ASIC development.

Professional & Community Service

  • Program Committee Member, PyTorch Conference 2025
    • Compilers & Kernels track. Reviewed proposals focused on compiler integration, deployment, and performance optimization topics.
    • Event site: PyTorch Conference 2025

Education

M.S. in Information Science and Technology
Osaka Institute of Technology, Japan — Apr 2015 – Mar 2017

B.A. in Information Science and Technology
Osaka Institute of Technology, Japan — Apr 2011 – Mar 2015

Publications (Selected)

Patents

Media Coverage

Awards & Recognition

  • PyTorch Community Awards 2025 — Nominee
  • PyTorch Community Awards 2024 — PyTorch Innovator Nominee