🍔 美國速食 AI 科技領導業者｜Senior Machine Learning Engineer｜Hybrid x 高年薪-Ting

Key Responsibilities

• Drive iteration across our AI/ML stack — object detection and tracking, person/vehicle Re-ID, and video understanding running on edge.
• Build efficient algorithms for resource-constrained hardware — implement lightweight architectures and optimization techniques within latency and compute budgets.
• Turn business asks into algorithmic problems — design the metrics, run experiments, and drive each iteration with ablation studies and error analysis.
• Partner with product engineers to ship and monitor ML systems in production — deployment paths and feedback loops that catch data drift and failure modes.
• Improve data and labeling quality — sampling strategy, annotation guidelines, and tooling that keeps a long-lived dataset healthy.

Requirements

Requirements

• 5+ years building production ML systems, with deep hands-on experience in computer vision, especially in object detection and multi-object tracking.
• Strong ML/DL fundamentals — statistics, classical methods, and modern deep learning, with a clear grasp of model internals, training dynamics, and common failure modes.
• Python and DL framework experience — PyTorch or TensorFlow, including custom training loops, distributed training, and end-to-end model debugging.
• A track record of driving research independently — picking the metric, designing the experiment plan, and interpreting noisy results honestly.
• Production ML sensibility — comfortable with inference engine (ONNX, OpenVINO, TensorRT), performance profiling, and porting models to real hardware.

Bonus points

• MLOps experience — experiment tracking, model and data versioning, reproducible training workflows (MLflow, DVC, or similar).
• ML pipeline / workflow orchestration — Dagster or similar tooling for training, evaluation, and deployment pipelines.
• LLM, RAG, or agentic AI experience — fine-tuning (LoRA/PEFT), retrieval pipelines (vector stores, rerankers), agent frameworks (LangChain or similar), or vision-language