AI Algorithm Engineer (AI Training/Inference Acceleration)

Company Overview

Our client is a globally recognized leader in information and communications technology infrastructure and smart devices, delivering comprehensive AI solutions across carriers, enterprises, governments, and individual consumers worldwide.

Role Overview

We are seeking an experienced AI Algorithm Engineer specialized in training and inference acceleration to join the team in Singapore. This role involves spearheading research, development, and implementation of advanced AI acceleration algorithms optimized for next-generation AI hardware.

Key Responsibilities

• Lead innovation in AI training and inference acceleration algorithms, focusing on Agentic AI and Multimodal domains to enhance compute efficiency on cutting-edge AI architectures.
• Manage the full development lifecycle of acceleration algorithms within proprietary AI frameworks and libraries, ensuring smooth integration and continuous performance improvement based on real-world data.
• Provide strategic technical insights within training and inference, identifying trends like long-sequence modeling and sparsity to future-proof and maintain competitive advantage.

Candidate Requirements

• Master's or PhD in Computer Science, Artificial Intelligence, Electronic Engineering, or related fields.
• In-depth knowledge of Large Language Models and Mixture of Experts architectures, with demonstrated experience in training, fine-tuning, or deploying large-scale AI models.
• Proficiency in AI acceleration technologies including zero-redundancy optimizers, distributed parallelism methods (TP, PP, SP, VP, DP), communication compression, and memory use optimization.
• Expertise in high-performance attention mechanisms, KV-cache compression, quantization techniques, and sparsity-aware accelerations is essential.
• Strong programming skills in Python coupled with extensive experience in prominent deep learning frameworks and hardware-accelerated libraries; experience in hardware-level kernel tuning is highly advantageous.
• Comprehensive understanding of hardware-software co-design principles, familiarity with hardware constraints like memory bandwidth and compute cycles, and successful performance optimization for models exceeding 100 billion parameters.
• Demonstrated ability to optimize AI models in distributed and heterogeneous systems by effectively leveraging hardware specifics for algorithmic enhancement.

Additional Information

Interested applicants should note that only shortlisted candidates will be contacted. By applying, candidates consent to the use and disclosure of their personal information as outlined in PERSOL Singapore Pte Ltd's Privacy Policy.

EA License No.: 01C4394