Michael Horgan
phone: 650-804-2651 Linkedin
Software Development — Machine Learning Research and Development — Real-time Signal Processing
Personal Projects
An LLM-inspired ChessBot Trained on Human Chess Games
I trained a transformer neural network on human chess games from the Lichess database to play chess in the style of a human player. The training procedure mirrors that of the "pre-training" phase for a Large Language Model.
2025
Modeling Chaotic Dynamics with Deep Learning: A Case Study on the Lorenz Attractor
I investigated the ability of a common neural forecasting architecture (NHITS) to model a chaotic dynamical system (the Lorenz Attractor). I trained the model on a PaperSpace VM using Pytorch and two RTX-5000 GPUs. (Github repo)
2024
A Scheduling Web App for Freedive SuperHome
I taught myself web programming and created a scheduling application pro bono for my favorite freediving school in the Philippines (Github repo)
2023
Competitive Freediving National Record
After dedicated daily training for approximately one year, I set a US national record in the "Dynamic No Fins" (DNF) freediving – pool discipline. (See my AIDA profile)
2023
Experience
Dolby Laboratories — San Francisco, CA
Staff Software Engineer, Applied AI Group
  • research and develop deep neural networks for noise reduction, speech separation, and speech synthesis, written in Pytorch and trained on NVIDIA GPU clusters
  • one of three research engineers who developed an original DNN-based speech separation network from the ground up; used by BlueJeans Video Conferencing for noise suppression
2018 - 2021
Staff Software Engineer, Consumer Entertainment Group
  • design, implement, and test real‑time spatial audio processing modules, written in C/C++, used for cinematic content creation
  • research and develop data-driven quality assurance tools for evaluating spatial audio content generation algorithms, written in Python
  • research and develop deep neural networks for speech synthesis, written in Pytorch and trained on NVIDIA GPU clusters
2015 - 2018
Senior Software Engineer, Advanced Technology Group
  • design, implement, and test software libraries for real-time spatial audio processing, perceptual audio coding, and loudness management, written in C/C++ for Intel, ARM, and Texas Instruments platforms
2011 - 2015
Applications Engineer, DSP Implementation Group
  • provide technical guidance to dsp core designers and integrated circuit developers implementing Dolby’s IP on proprietary architectures
  • validate and issue design approvals of licensees’ core designs, IC implementations, and professional audio/video processing products
2008 - 2011
Publications
B. Kadioglu, M. Horgan, X. Liu, J. Pons, D. Darcy, V. Kumar, "An Empirical Study of Conv-TasNet" ICASSP 2020, arXiv:2002.08688
C. Zhou, M. Horgan, V. Kumar, C. Vasco, D. Darcy, “Voice Conversion with Conditional SampleRNN” Interspeech 2018, arXiv:1808.08311
Patents
X. Liu, M. Horgan, R. Fejgin, P . Holmburg. Deep Learning Based Speech Enhancement. WO 2022/094293 A1
C. Zhou, M. Horgan, V. Kumar, C. Vasco, J. Morales. Speech Style Transfer. US 11,538,455 B2
C. Zhou, X. Liu, M. Horgan, V. Kumar. Systems and Methods for Adapting Human Speaker Embeddings in Speech Synthesis. US 2022/0335925 A1
X. Li, G. Cengarle, Q. Bin, M. Horgan. Generating Channel and Object-based Audio from Channel-based Audio. WO 2023/076039 A1
B. Kadioglu, M. Horgan, J. Pons Puig, X. Liu. Deep Source Separation Architecture. WO 2021/081002 A1
Education
Stanford University
M.A. in Music, Science, and Technology
2008
Stanford University
B.A. in Music, Science, and Technology
2008