AI Engineer (Fluent in Mandarin & English)

Overview

In this role you will own the end-to-end production lifecycle of AI, from training and fine-tuning domain LLMs to building scalable inference pipelines. You will write clean, Python-based code and deliver reliable, low-latency AI solutions for voice and text applications. You’ll work with cross-functional partners to translate business needs into robust ML systems, embedded in containerized microservices. This is a hands-on role at scale, combining MLOps, performance engineering, and AI delivery to impact real-world use cases.

Pay / Benefits

• Competitive salary
• pension scheme
• discretionary bonus
• 25 days annual leave + 5
• hybrid working options
• Private Medical cover

Responsibilities

• Lead domain-specific LLM adaptation using LoRA, QLoRA, and PEFT to balance performance and resources
• Architect and optimize inference pipelines to reduce Time to First Token and increase throughput (quantization, caching, batching)
• Build and maintain real-time AI pipelines using WebSockets and SSE for voice (ASR/TTS) and text apps
• Deploy and orchestrate models within Docker/Kubernetes-based microservices with monitoring, security, and scalability
• Collaborate with Business Analysts and stakeholders to align commercial needs with technical execution

Key requirements

• 5+ years in AI/ML engineering with production-ready model deployment experience
• Deep proficiency in Python with clean coding, modular design, and testing
• Hands-on experience with LLM training cycles, PEFT, and prompt engineering
• Experience with high-performance inference servers (vLLM, TGI, or Triton) and GPU deployment optimization
• Linux-based environment expertise and proficiency with containerized workflows and CI/CD
• Experience building Retrieval-Augmented Generation (RAG) systems, including vector databases and semantic search

• collaboration with cross-functional teams
• effective communication with stakeholders
• problem-solving under performance constraints
• LoRA, QLoRA, PEFT for model fine-tuning
• inference optimization (TTFT, throughput)
• quantization, caching strategies, and efficient batching

…