Immersive Audio AI Researcher

TikTok; Sydney; On-site; AUD 120,000 - 180,000; Full time

Job summary

TikTok's Audio Innovation team in Sydney is seeking a research-focused engineer to push immersive audio technology across short video, live streaming, VR/AR, and content tools. You will develop self-developed algorithms for immersive audio understanding and processing.

You will contribute to international standards bodies, publish proposals, and work with ML/DL models, Kaldi/TensorFlow/PyTorch, and Python/C++. A strong background in 3D spatial audio is preferred.

Qualifications

• Master's degree or higher in electronic information, computer science, multimedia, ML/AI.
• Experience in 3D Spatial Audio rendering or GAN/Diffusion audio processing, voiceprint/wake-up, sound event detection, ASR or NLP.
• Proficiency in Python and C++, familiarity with Kaldi/TensorFlow/PyTorch and large-scale model training.

Responsibilities

• Research and implement the department's self-developed immersive intelligent audio understanding and processing algorithm system.
• Develop and optimize immersive audio codec and processing for short video, LIVE, VR/MR and related scenarios.
• Follow up on domestic and international audio standards; write high-quality technical proposals and contribute to standards work.

Skills

Data structures & algorithms

Deep learning model design

Python programming

C++ programming

Model training on large datasets

Education

Master's degree or above in electronic information, computer, multimedia, ML/AI

Tools

Kaldi

TensorFlow

PyTorch

Job description

TikTok's Audio Innovation team in Sydney is seeking a research-focused engineer to push immersive audio technology across short video, live streaming, VR/AR, and content tools. You will develop self-developed algorithms for immersive audio understanding and processing.

You will contribute to international standards bodies, publish proposals, and work with ML/DL models, Kaldi/TensorFlow/PyTorch, and Python/C++. A strong background in 3D spatial audio is preferred.