Edge AI & Generative Modeling Engineer

MediaTek; Singapore; On-site; SGD 90,000 - 150,000; Full time

Job summary

MediaTek in Singapore seeks an AI software engineer to advance generative AI and edge AI for mobile and embedded devices. You will work on transformer-based models, RNN/SSM architectures, and on-device NLP and vision tasks.

You will optimize models for constrained hardware, explore quantization and pruning, and contribute to design automation and RL/on-device learning. Proficiency in Python, PyTorch, and C/C++ is required, with hands-on experience in TFLite or similar frameworks.

Qualifications

• Strong understanding of AI models (especially transformer and LLMs) — graph structure, operators and inference behaviour.
• Familiar with on-device frameworks for mobile/embedded devices: TFLite, ExecuTorch
• Required programming languages: Python, PyTorch, C/C++
• Knowledge of deep learning and generative AI algorithm/model development, optimization and training methodology preferred
• Experience developing deep learning and generative AI algorithms for mobile/embedded devices is an advantage

Responsibilities

• Generative AI: language and multi-modal model algorithm development
• Transformer-based architecture (LLaMa etc.)
• RNN and SSM-based architecture (Mamba, Linear Attention, etc.)
• AI vision / voice / NLP algorithm development
• Edge AI (APU) algorithm development and optimization
• Neural network optimization (quantization / pruning / low-bit techniques)
• Neural network design automation
• Advanced AI algorithm: reinforcement learning (RL) and on-device learning

Skills

Transformer models

LLMs

On-device ML

Python

PyTorch

C/C++

Deep learning

Model optimization

Mobile/embedded

Tools

TFLite

ExecuTorch

Job description

MediaTek in Singapore seeks an AI software engineer to advance generative AI and edge AI for mobile and embedded devices. You will work on transformer-based models, RNN/SSM architectures, and on-device NLP and vision tasks.

You will optimize models for constrained hardware, explore quantization and pruning, and contribute to design automation and RL/on-device learning. Proficiency in Python, PyTorch, and C/C++ is required, with hands-on experience in TFLite or similar frameworks.