Cheng-i Wang 汪正一 汪正一（汪六）Cheng-i Wang
Deep Learning Engineer at AudioShakeAudioShake 深度學習工程師
Cheng-i Wang (nicknamed “Wang Six”) is a Deep Learning Engineer at AudioShake, where he develops audio source separation technology and brings deep learning models into production. Before joining AudioShake, he spent four years as an Audio Software Engineer at Smule, building data-driven solutions for karaoke at consumer scale and architecting large-scale audio processing pipelines and ML infrastructure.
His training combines engineering and music: a PhD in Computer Music from UC San Diego, an MM in Music Technology from NYU, and a BBA in Finance from National Taiwan University. He was a recipient of the Taiwan Ministry of Education Government Scholarship for Overseas Study (Applied Music) in 2009.
Before moving abroad for graduate studies, he was an active working musician in Taiwan: bassist and bandleader of Apay, bassist for live concerts by Sandee Chan and Waa Wei, and composer, arranger and performing music director for Tainaner Ensemble’s stage production Castrated Chicken.
汪正一（汪六），AudioShake 深度學習工程師，做 AI 音訊分離。加入 AudioShake 之前，於 Smule 擔任音訊軟體工程師四年，負責卡拉 OK 應用的音訊處理與機器學習系統。
UCSD 電腦音樂博士、NYU 音樂科技碩士、台大財金學士，2009 年教育部公費留學獎學金（應用音樂）。
出國前是職業 bass 手——阿霈樂團（Apay）團長、陳珊妮與魏如萱演唱會樂手、臺南人劇團《閹雞》配樂兼台南場指揮。
Updates近況
- Spoke at Taiwan Multimedia Tech meetup #12 in Taipei on “What LLMs Left Untouched in Audio”, a look at the structural split inside audio AI.於台北 Taiwan Multimedia Tech #12 分享「LLM 沒吃掉的那塊 Audio」，談 audio AI 的結構性分裂。
- AudioShake launched Dialogue RT, dialogue isolation at 11 ms end-to-end for live broadcast. I worked on getting the low-latency separation to ship.AudioShake 發表 Dialogue RT，11 毫秒端到端的即時人聲分離，用於直播。我負責低延遲分離的部分。
- Gave a tech talk at KKCompany (KKBOX) in Taipei on AI source separation and how it differs from generative models.於台北 KKCompany（KKBOX）分享 AI 音訊分離技術，以及它與生成式模型的差異。
- Joined AudioShake as a Deep Learning Engineer.加入 AudioShake 擔任深度學習工程師。
- Co-authored “ToNet: Tone-Octave Network for Singing Melody Extraction from Polyphonic Music” at ICASSP 2022.合著論文 “ToNet: Tone-Octave Network for Singing Melody Extraction from Polyphonic Music” 發表於 ICASSP 2022。
- Co-authored “Deep Autotuner” at ICASSP 2020 and the TISMIR survey on audio-based music structure analysis.合著 “Deep Autotuner”（ICASSP 2020）與 TISMIR 音樂結構分析回顧論文。
- US patent 10,446,123, “Intuitive music visualization using efficient structural segmentation”, issued (with Adobe Research).美國專利 10,446,123「Intuitive music visualization using efficient structural segmentation」核准（與 Adobe Research 合作）。
- Joined Smule as an Audio Software Engineer, and presented “Singing Style Investigation by Residual Siamese Convolutional Neural Networks” at ICASSP 2018 in Calgary.加入 Smule 擔任音訊軟體工程師，並於 ICASSP 2018（卡加利）發表 “Singing Style Investigation by Residual Siamese Convolutional Neural Networks”。
- Graduated from UC San Diego with a PhD in Music (Computer Music).取得 UC San Diego 音樂博士（電腦音樂）學位。