AI Systems Engineer: End-to-End LLM Production & Impact

Hyphen Connect Limited; Hong Kong; On-site; HKD 900,000 - 1,200,000; Full time

Job summary

Hyphen Connect Limited is seeking an AI Engineer to bridge cutting-edge model capabilities with tangible product value. You will own the AI lifecycle end-to-end, shaping model behavior and engineering surrounding systems for robust, scalable performance at real-world scale.

You will collaborate with product and engineering teams to ship production-ready features, orchestrate prompts and tools, and ensure deterministic behavior across pipelines while driving continuous improvement in accuracy and

Qualifications

• Strong foundation in machine learning principles and neural networks.
• Experience deploying large models to production.
• Proficient in writing clean, maintainable production code.

Responsibilities

• End-to-End Feature Delivery: Own the pipeline from raw model to user experience, shipping production-ready AI features.
• Agentic Design & Orchestration: Architect prompts, tool-use, memory retrieval, and autonomous agent workflows.
• Behavior Standardization: Turn raw model outputs into deterministic, dependable system behaviors.
• Full-Stack Troubleshooting: Debug issues across models, orchestration, infra, and UX.
• Performance Engineering: Optimize architecture for low latency, cost-efficiency, and uptime.
• Data-Driven Evaluation: Build lightweight frameworks to measure real-world model accuracy.
• Collaborative Problem Solving: Work with product and engineering teams to translate challenges into code and systems.

Skills

Python

Machine Learning

System design

Tools

PyTorch

JAX

OpenAI APIs

LLaMA

Qwen

vLLM

Vector databases

Job description

Hyphen Connect Limited is seeking an AI Engineer to bridge cutting-edge model capabilities with tangible product value. You will own the AI lifecycle end-to-end, shaping model behavior and engineering surrounding systems for robust, scalable performance at real-world scale.

You will collaborate with product and engineering teams to ship production-ready features, orchestrate prompts and tools, and ensure deterministic behavior across pipelines while driving continuous improvement in accuracy and