Hello there!

I'm Kai-Yin Hung - a software engineer who keeps picking venture teams over safer paths. These days I build ML compiler and runtime stacks for low power AI SoCs at Upbeat Technology, a Series B startup in Taipei.

If you’d like the structured view, here’s my resume .

Below is the narrative version—why I keep betting on fast-moving ventures and what I’ve shipped along the way.

Kai-Yin Hung

Biography

I grew up in Taichung, a city that sparked my curiosity about how things work and what lies beyond what I could see every day. That curiosity deepened when I watched Silicon Valley (2014–2019), a show that captured the mix of chaos and creativity in startup life. Watching it made me believe that with enough persistence and imagination—like Richard—I could also build something meaningful from an idea. That moment was when my interest in technology and startups really began.

While Silicon Valley represented the dream, I’ve always believed Taiwan has its own potential. Our hardware industry is world-class, but our software side still has room to grow. I wanted to help bridge that gap and be part of Taiwan’s own startup story. I knew it would take more than technical skills—it would take learning how to build real products that solve real problems. Since then, every step I’ve taken has been guided by that goal.

I earned my B.Eng. (2022) and M.Eng. (2024) in Electrical and Computer Engineering at National Yang Ming Chiao Tung University. Those years gave me a strong foundation in both hardware and software, but more importantly, they taught me how to keep learning and stay patient. I discovered that my biggest strength is debugging—slowing down, looking deeply, and making sense of messy problems. That mindset still guides how I work today.

During my internships at GallopWave and ITRI, I got my first chance to connect theory with real-world systems. I worked on projects that combined sensors, algorithms, and unpredictable data from real traffic. It was the first time I saw how ideas turn into working systems—and how good engineering is less about perfect theory and more about making things work reliably.

At InQuartik, a patent-intelligence SaaS company, I moved from robotics to software. It was the first time I worked on products that real users depended on. I learned about writing maintainable code, setting up CI/CD, and collaborating with product teams. I also saw how fast market needs can change what we build and how we build it. That experience helped me see software not just as code, but as a way to deliver value and adapt quickly.

After finishing my master’s, I joined Upbeat Technology—a venture team that bridges hardware and software. We design RISC-V SoCs with dedicated AI accelerators, and I focus on the low-level software that powers them. Since I joined, the team has grown quickly, the product has moved from R&D to production, and revenue has multiplied. Beyond the technical work, I’ve learned how a company grows—how different people and roles come together to build something real. It’s helped me see what it truly takes to turn an idea into a company.

Career Timeline

Roles that shaped how I build software — from research labs to real-world impact.

  1. Software Engineer @ Upbeat Technology
    Jul 2024 → Present (1y 6m)
    • Built AI from neural network to chip.
  2. Software Engineer Intern @ InQuartik
    Jan 2022 → Dec 2023 (1y 11m)
    • Shipped multi-tenant Vue + Spring MVC features and automated HubSpot & Apollo CRM sync, reducing manual marketing workload by 60%.
  3. Software Engineer Intern @ Industrial Technology Research Institute (ITRI)
    Oct 2021 → Oct 2022 (1y)
    • Designed motion-forecasting networks and automated 3D point-cloud pipelines for L4/L5 autonomous driving projects.
  4. Software Engineer Intern @ GallopWave
    Jun 2021 → Sep 2021 (3m)
    • Implemented C++ testing harnesses for embedded visual–inertial odometry systems across IMU/Camera/GPS configurations.

Selected Builds

A few products and experiments that capture what I like to ship.

  1. House168

    2024 → Now

    Website serving 4,000+ real estate agents & agencies; keep shipping front-end solutions.

    • Full-stack
    • Data Engineering
  2. doudou.jobs

    2025

    Shipping full-stack solutions for doudou, the Employer of Record (EOR) brand under HRNetGroup.

    • Full-stack
    • System Design
    • ERP
  3. Motion Prediction on Waymo Open Motion Dataset

    2022

    Ranked 8th globally for multi-agent trajectory forecasting for autonomous vehicles. Honored to be part of the team with my senior lab mentor, which greatly enriched my research experience.

    • Autonomous Driving
    • Waymo
    • Deep Learning
  4. Multi-Modal Motion Prediction using Temporal Ensembling with Learning-based Aggregation

    2024

    Open source my IROS 2024 paper codebase on multi-modal motion prediction for autonomous vehicles.

    • Computer Vision
    • Deep Learning
    • PyTorch
    • Motion Prediction
    • L4/L5 Autonomous Vehicles
    • Argoverse 2 Dataset