[台北] Machine Learning Engineer
Posted Updated
Description
About the Role
- We are seeking a Machine Learning Engineer with strong engineering fundamentals to build production-ready ML systems.
- The problem space is not fixed and may evolve over time, including (but not limited to) computer vision, tracking, LLM/VLM, or other applied ML domains.
- Rather than focusing on a single research direction, this role emphasizes practical execution, system integration, and long-term operability.
- You will work closely with software, platform, and infrastructure teams to ensure ML components are efficient, reliable, and maintainable in real production environments.
Responsibilities
- Model Development & Evaluation
- Train, fine-tune, and evaluate machine learning models across different problem domains
- Understand data characteristics, evaluation metrics, and real-world failure modes
- Algorithm Understanding & Application
- Apply and adapt core algorithms commonly used in applied ML systems
- Understand and work with fundamental methods such as state estimation, data association, and optimization when relevant
- Inference & Performance Optimization
- Optimize inference latency, throughput, and resource utilization
- Make informed trade-offs between model quality, performance, and system constraints
- Productionization & Integration
- Package models into reusable tools, libraries, or services
- Integrate ML components into existing systems with clear configuration and runtime controls
- Support deployment and validation in production (cloud and/or edge)
- System Debugging & Collaboration
- Diagnose issues across model behavior, runtime performance, and system stability
- Collaborate with cross-functional teams and clearly communicate technical constraints
Required Qualifications
- Proficiency in Python with solid software engineering practices
- Experience with at least one major ML framework (e.g., PyTorch, TensorFlow)
- Strong understanding of machine learning fundamentals
- Training, evaluation, generalization, and performance trade-offs
- Familiarity with core algorithms and mathematical concepts commonly used in applied ML systems
- e.g., Kalman filtering, assignment / matching problems (Hungarian algorithm), optimization basics
- Ability to move ML work beyond experimentation into reliable, usable systems
- Comfortable working with changing requirements and problem domains