Skip to content

About Me

Personal Mission Statement

"Advancing human language technology and machine learning to deliver practical, impactful, and measurable improvements in everyday life."

Career Highlights

  • Former Speech Architect of Voci’s speech recognition engine (2012-2023)
  • Former maintainer of the open-source CMU Sphinx project (2004–2006)
  • Early team member at two startups:
  • ScanScout (#4 employee), acquired by Tremor Video in 2010
  • Voci (#9 employee), later acquired by Medallia
  • Research staff roles at Raytheon BBN, SpeechWorks (now part of Nuance), ScanScout, and Voci
  • Co-author of a best paper at a prestigious international conference

Skills

Programming

  • Languages: C, Perl, Python, C++, Java
  • Also familiar with web and iOS programming

Conversational AI

My expertise spans the full stack of Conversational AI technologies, including automatic speech recognition (ASR), large language models (LLMs), and text-to-speech (TTS).

  • ASR: Proficient across multiple generations of speech recognition systems, from traditional HMM-based models and HMM-DNN hybrids to modern foundation model–based approaches
  • TTS: Experienced in fine-tuning state-of-the-art TTS foundation models for custom voice synthesis
  • LLM: Deep experience with large language models, including prompt engineering, fine-tuning, and applying them to real-world NLP tasks.

🗣️ Speech Recognition

  • System architecture (decoder + trainer)
  • Very fast speech recognition
  • Keyword spotting
  • Classification tasks: topic, language, emotion, gender
  • Robust speech recognition

Toolkits used:
PyTorch, kaldi, Sphinx (2, 3, 4, PocketSphinx), HTK, Julius, SpeechWorks (pre-OSR 2.0), Byblos, CMULM, SRILM, MITLM, KenLM

Deep Learning

  • Large Language Model: Application of LLM on conversational systems, fine-tuning and RAG
  • Speech Recognition: Low-level source-code experience with multiple deep learning toolkits
  • Speech Synthesis: Fine-tuning CosyVoice2 for new language and accent correction.
  • Administration: Setup and deployment of deep learning tools

Toolkits:
PyTorch, Tensorflow.

Soft Skills

  • Startup experience and MVP building
  • Speech applications and analytics
  • Business use cases of speech recognition and machine learning (especially in startup contexts)
  • Expertise in open-source speech recognition ecosystems