LINE Careers

We are looking for a savvy ML Engineer with technical backgrounds in ML system development. LINE is being rebuilt around Large Language Models (LLMs), Retrieval-Augmented Generation (RAG) and AI Agents that already reach 20M+ users in Taiwan. As a core ML Engineer you will design, train, optimize and serve models that power LINE TODAY, Stickers, Shopping, VOOM and yet-to-be-imagined products.

 

Responsibilities

• End-to-end ownership of ML systems: problem framing, data collection, feature engineering, model training, evaluation, on-device / cloud deployment and post-launch monitoring.

• Fine-tune, distill and serve LLMs using DeepSpeed/FSDP/LoRA, with advanced techniques such as RLHF, quantization, pruning and knowledge distillation.

• Build LLM-powered agents (tool calling, multi-step reasoning, planning) and RAG pipelines that integrate search, vector databases and external APIs.

• Design + implement distributed training jobs and low-latency inference stacks (Triton, TensorRT).

• Set up and own CI/CD & LLMOps workflows—model registry, canary release, drift & bias detection, automated rollback.

• Drive online experimentation: define metrics, run large-scale A/B tests, apply sequential testing, and translate results into business impact.

• Build and own an LLM/agent observability & evaluation platform: implement offline/online eval harnesses (prompt, LLM-as-judge, human-in-the-loop); instrument agent telemetry & tracing, prompt versioning; detect hallucination/toxicity.

• Collaborate with product, backend, mobile and infra teams; present research and best practices in internal tech talks and external conferences.

 

Qualifications

• 3 years of post-graduate industry or research experience shipping ML/DL/AI systems.

• Deep understanding of modern ML algorithms (optimization, regularization, ensemble methods, sequence modeling, attention).

• Hands-on production experience with PyTorch/TensorFlow including custom ops, mixed-precision training and gradient checkpointing.

• Proven track record of deploying models and monitoring them in production.

• Fluency in statistical inference & experimental design; ability to interpret and communicate results to both technical and non-technical stakeholders.

• Excellent written and spoken English; comfortable authoring design docs and presenting to leadership.

 

Interview Process

• CV Screening
• On-line Assessment
• On-line Interview (if assessment passed)