Shawn Song

Sonilo is an AI video-to-music platform that automatically generates perfectly synchronized soundtracks for your videos. By analyzing visuals, pacing, and emotion, Sonilo creates music that fits your content frame-by-frame—no editing or manual timing needed. Built for creators, filmmakers, and creative tools, it delivers fast, high-quality results and integrates seamlessly into modern video workflows.

Leading a 50+ person team across research and engineering, focused on advancing state-of-the-art AI music generation. Owned the full-stack technology strategy for high-fidelity and controllable music generation, including model design, data systems, evaluation, and scalable training and inference infrastructure. Scaled Mureka.ai to $20M+ ARR, positioning it among the leading AI music generation platforms globally. Representative Work: Musicot: Analyzable Chain-of-Musical-Thought Prompting for High-Fidelity Music Generation

Leading a 50 members, including music generation (Mureka.ai), world model, multimodal understanding across algorithms, data, evaluation, AI infrastructure, and engineering. Focused on building the world’s best AI-powered song generator, driving innovation from cutting-edge research to large-scale production. Representative Work: Matrix-game 2.0: An open-source real-time and streaming interactive world model (https://arxiv.org/abs/2508.13009) Skywork UniPic 3.0: Unified Multi-Image Composition via Sequence Modeling (https://arxiv.org/abs/2601.15664) Skywork-R1V4: Toward Agentic Multimodal Intelligence through Interleaved Thinking with Images and DeepResearch (https://arxiv.org/pdf/2512.02395)

Tech lead and staff research engineer at ByteDance Seed Audio/Music; Leading the TikTok Multimodal Team to conduct cutting-edge tech in large-scale multimodal learning, Large Language Model (LLM) and end-to-end music generation.

Tech lead and staff research engineer at ByteDance Seed Audio/Music; Leading the TikTok Multimodal Team to conduct cutting-edge tech in large-scale multimodal learning, Large Language Model (LLM) and end-to-end music generation.

• Designed and enhanced multi-engine Speech Recognition (ASR), Natural Language Processing (NLP), and Natural Language Understanding (NLU) modules for Bosch's in-car voice assistant. • Developed intent detection and slot filling systems, improving user interaction for millions of users. • Collaborated with cross-functional teams to ensure seamless integration of voice technology into automotive applications.

Talented Student Program (Union Class of Electrical Eng)