On-Device Quantization Engineer for Large Models

XG TECH PTE. LTD.; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

XG Tech PTE. LTD. in Singapore seeks an experienced Large Model Quantization Algorithm Engineer to develop quantization and model compression for LLMs, VLMs, and video generation models, optimizing accuracy, memory, and on-device performance.

You will collaborate with algorithm, compiler, and hardware teams to deploy efficient AI inference into production, adapt models across NPU/CPU platforms, and build automation tools for quantization workflows.

Qualifications

• Bachelor’s degree or higher in a quantitative field (Engineering, CS, Math).
• Familiarity with LLM/VLM quantization and deployment optimization.
• Hands-on PyTorch QAT development is advantageous.
• Strong Python and C++ programming skills.

Responsibilities

• Develop and optimize quantization algorithms for LLMs, VLMs, and video generation models (PTQ, QAT, and related techniques).
• Design and evaluate quantization schemes balancing accuracy, performance, and memory.
• Perform calibration and error analysis for accuracy degradation and optimization.
• Optimize edge/on-device inference (Prefill/Decode acceleration, KV Cache, fusion, weight compression, memory).
• Adapt and deploy models across heterogeneous hardware (NPU/CPU); collaborate on model conversion and engine tuning.
• Develop quantization toolkits, workflows, accuracy evaluation, and debugging tools.
• Collaborate with model/architecture teams on quantization-friendly architectures and training strategies.
• Track and apply emerging quantization research to production.

Skills

LLM/VLM algorithms

QAT (PyTorch)

Python

C++

Edge deployment

Education

Bachelor’s degree in Electronics/CS/Quantitative field

Tools

PyTorch

Job description

XG Tech PTE. LTD. in Singapore seeks an experienced Large Model Quantization Algorithm Engineer to develop quantization and model compression for LLMs, VLMs, and video generation models, optimizing accuracy, memory, and on-device performance.

You will collaborate with algorithm, compiler, and hardware teams to deploy efficient AI inference into production, adapt models across NPU/CPU platforms, and build automation tools for quantization workflows.