Large Model Quantization Algorithm Engineer

KINGFISHER RECRUITMENT (SINGAPORE) PTE. LTD.; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

KINGFISHER RECRUITMENT (SINGAPORE) PTE. LTD. partners with an innovative semiconductor company building its first overseas R&D team in Singapore. The role focuses on quantization and model compression for LLMs, VLMs, and video generation models across NPU, GPU and CPU platforms.

You will work with algorithm, compiler and hardware teams to optimize accuracy, memory efficiency and inference performance while developing supporting toolkits for automated quantization workflows and debugging.

Qualifications

• Bachelor's degree in a quantitative discipline; required for entry
• Familiarity with LLM/VLM algorithms and deployment optimisation techniques
• Strong Python and C++ programming skills
• Hands-on PyTorch experience is a plus

Responsibilities

• Develop and optimize model quantization algorithms for LLMs, VLMs, and video generation models.
• Balance accuracy, memory efficiency and inference performance across NPU, GPU and CPU.
• Perform calibration and error analysis to identify sources of accuracy degradation.
• Collaborate with algorithm, compiler and hardware teams on deployment and optimization.
• Build quantization tooling and automated workflows for evaluation and debugging.
• Explore emerging quantization research and translate techniques into production solutions.

Skills

Python

C++

Quantization

Model compression

PTQ/QAT concepts

Education

Bachelor's degree in a quantitative discipline

Tools

PyTorch

Job description

Large Model Quantization Algorithm Engineer | LLM / VLM / Edge AI

Our client is an innovative semiconductor company expanding its international R&D footprint, developing advanced computing platforms for intelligent vehicles and edge AI applications. As part of this expansion, the company is building its first overseas R&D team in Singapore and is seeking a Large Model Quantization Algorithm Engineer.

Role Overview

You will develop quantization and model compression algorithms for large language models (LLMs), vision-language models (VLMs), and video generation models. Working closely with algorithm, compiler and hardware teams, you will optimise model accuracy, memory efficiency and inference performance for deployment across NPU, GPU and CPU platforms.

Key Responsibilities

• Develop and optimise model quantization algorithms, including post-training quantization (PTQ), quantization-aware training (QAT), and related compression techniques.
• Design and evaluate quantization schemes that balance model accuracy, inference performance and memory efficiency.
• Perform quantization calibration and error analysis to identify and address sources of accuracy degradation.
• Optimise on-device inference through prefill/decode acceleration, KV cache management, operator fusion, weight compression and memory optimisation.
• Adapt and deploy models on heterogeneous hardware platforms, collaborating with compiler teams on model conversion, engine compilation and performance tuning.
• Develop model optimisation toolkits, including automated quantization workflows, accuracy evaluation and visualisation/debugging tools.
• Collaborate with model and architecture teams on quantization-friendly model architectures, training strategies and inference optimisation.
• Evaluate emerging research in quantization, compression, sparsity and efficient inference, and translate relevant techniques into production solutions.

Requirements

• Bachelor's degree or above in Electronic Engineering, Computer Science, Automation, Operations Research, Statistics, Mathematics or a related quantitative discipline.
• Familiarity with LLM/VLM algorithms and deployment optimisation techniques, including quantization, sparsity, pruning and inference acceleration frameworks.
• Strong Python and C++ programming skills.
• Hands-on experience developing quantization-aware training solutions with PyTorch would be an advantage.