👋 Hi there! I'm Chien.
I'm an engineer at the intersection of machine learning, infrastructure, and full-stack development. At Aignostics, a medical AI startup, I build scalable ML pipelines and developer tooling that power drug discovery research. My background spans LLM fine-tuning at NIO, distributed database research at IBM, and computer vision work published at CVPR and ICML — always focused on real-world, measurable impact. When I'm not coding, I enjoy cooking and playing badminton.
Hi there! I've been trained on Chien's professional background and documentation.
Feel free to ask me any questions about Chien!
Experience
- JULY 2024 - FEBRUARY 2026
MLOps Engineer - GCP @ Aignostics
Developed ML pipelines for early-stage lung cancer discovery.
- Redesigned ML training pipelines on GKE, cutting training time by 90%, which enabled the data science team to run significantly more experiments at lower cost while keeping critical research milestones on schedule.
- Built an internal GPU provisioning AI platform on GCP that eliminated 18+ hours/day of wait time for the data science team, removing a major bottleneck and accelerating model development cycles.
- Cut Docker image build time by 90%, shrinking CI/CD cycles so engineers could test and deploy changes in minutes instead of hours, which improved overall release velocity.
- DECEMBER 2023 - MAY 2024
Software Engineer Intern - AWS @ NIO
Develop car AI assistant in NOMI team.
- Improved speech recognition accuracy by 35% over Microsoft's baseline by fine-tuning Whisper from HuggingFace, directly improving the reliability of NOMI in-car voice assistant, especially in multi-language use cases, e.g. "Navigate to Berlin Hauptbahnhof".
- Built an automated benchmarking leaderboard for ASR models with Airflow, freeing up 10+ hours/week previously spent on manual QA, which let the team focus on model improvements instead of evaluation overhead.
- OCTOBER 2022 - MAY 2023
Software Engineer Intern @ IBM
Designed an ML-based anomaly detection algorithm for PostgreSQL on Kubernetes as part of my Master's thesis — laying groundwork for catching malicious or anomalous database activity early, before it causes downtime or security incidents.
- Achieved 95% accuracy on spotting anomalous events with unsupervised learning models.
- Reduced 85% of metric space by feature engineering to retain relevant metrics.
- May 2022 - SEPTEMBER 2022
Student Research Assistant @ Universität Rostock
Analyzed fuel consumption of large cargo vessels to identify key cost drivers, supporting route optimization for a shipping company.
- Reduced fuel consumption prediction error by 60% by developing machine learning models to identify the primary factors driving fuel usage — giving the cargo company more reliable forecasts to optimize shipping routes and cut fuel costs.
- Built an interactive data visualization dashboard for exploratory data analysis, enabling the research team to quickly identify and communicate the biggest factors behind fuel consumption across the fleet.
- APRIL 2021 - SEPTEMBER 2021
Data Scientist Intern @ Teraki
Developed radar object detection models for autonomous driving system.
- Accelerated radar data preprocessing by 10x by rewriting the Python bottleneck in Cython, cutting training pipeline runtime from hours to minutes, which enabled the team to iterate on end-to-end radar object detection models for the first time.
- Developed end-to-end deep learning models in PyTorch for radar-based vehicle and pedestrian detection, deployed on edge hardware, contributing working perception models to Teraki's safety-focused ADAS stack.
- OCTOBER 2018 - JUNE 2020
Machine Learning Researcher @ ElsaLab
Researched machine learning approaches in image processing and reinforcement learning; designed, implemented, and evaluated experimental frameworks. The results are published in top conferences such as CVPR and ICML.
- Co-authored CVPR 2020 and ICML 2020 papers on visual odometry and reinforcement learning, contributing methods that advanced the lab's research on autonomous navigation and robot decision-making.
- Built distributed training infrastructure across GCP and on-prem GPUs, cutting experiment turnaround from weeks to days — the speed-up that made the lab's publication output possible.
- Implemented C++/ROS control systems to deploy trained models on physical robots, validating research results in real-world navigation conditions.
- MAY 2019 - AUGUST 2019
Student Research Assistant @ National Tsing Hua University
Led an industrial-academic collaboration project to develop an intelligent seeding machine for Malabar Chestnut.
- Designed image classification algorithms using OpenCV and Python to improve seed orientation detection and increase production rate by 90%.
- Implemented automation control systems for motors and pneumatic devices using C language and embedded systems.