AI Infrastructure Intern — LLM Optimization & Deployment

TikTok; Singapore; On-site; SGD 20,000 - 27,000; Full time

Job summary

TikTok is seeking interns to join the E-commerce Recommendation Infra team in Singapore. You will research and implement LLM-based recommendation foundations, and build optimization toolchains for large-scale deployment in a fast-growing e-commerce context.

Candidates should be pursuing a CS/CE/IS degree and have hands-on experience with C++/CUDA/Python and DL frameworks like PyTorch or TensorFlow. Prior exposure to Megatron-LM or DeepSpeed is a plus.

Qualifications

• Currently pursuing a Bachelor's or Master's degree in Computer Science, Computer Engineering, Information Systems, or a related technical discipline
• Solid programming skills in C++/CUDA/Python, with basic understanding of GPU architecture and distributed training
• Familiarity with deep learning frameworks (PyTorch, TensorFlow) and basic knowledge of LLMs

Responsibilities

• Research and implement LLM-based recommendation foundation model training and inference optimization technologies for e-commerce scenarios
• Build model optimization toolchains including quantization, pruning, and distillation for deep learning frameworks
• Design and implement distributed inference serving architectures supporting large-scale LLM deployment for recommendation use cases

Skills

C++/CUDA/Python

GPU architecture

Distributed training

Deep learning frameworks

LLM inference acceleration

Education

Bachelor’s or Master’s in CS/CE/IS

Tools

PyTorch

TensorFlow

Megatron-LM

DeepSpeed

TensorRT-LLM

Job description

TikTok is seeking interns to join the E-commerce Recommendation Infra team in Singapore. You will research and implement LLM-based recommendation foundations, and build optimization toolchains for large-scale deployment in a fast-growing e-commerce context.

Candidates should be pursuing a CS/CE/IS degree and have hands-on experience with C++/CUDA/Python and DL frameworks like PyTorch or TensorFlow. Prior exposure to Megatron-LM or DeepSpeed is a plus.