Skip to content

AI Portfolio

For the most part, working on machine learning means creating programs that work in real life.
In that regard, here are selected systems I've worked on over the years.
Noted work is bolded. Not all work is shown — feel free to contact me if you have questions on specific AI/LLM/deep-learning-related problems.

Automatic Speech Recognition / Speech-based ML Systems

Grad School Years (2000–2002)

  • Viterbi algorithm implementation that allows skipping exactly K-frames — featured in several academic papers
  • Pronunciation learning system, also known as PLACER

Speechworks (2002–2003)

  • Cantonese, Singaporean, and Australian English model training for Speechworks 6.5

Carnegie Mellon University (2003–2006)

  • Maintainer of CMUSphinx, especially Sphinx3 and SphinxTrain

ScanScout / Tremor (2006–2008)

  • Internal system known as “Content Analyzer” using speech recognition as a backend component

BBN Technologies (2009–2011)

  • Unsupervised topic/dialogue classifier based on PTM/Byblos(Hmm-based) segmentation
    Selected as one of the best papers at Interspeech 2010
  • British English model training

Voci Technologies (2012–2023)

  • Architect and Maintainer of Voci’s high-speed speech recognizer, also known as V-Blaze
  • Manager of a three-person team, responsible for language pack delivery
  • General Speech Advisor of the group
  • High-performance keyword spotter
  • Speech-based gender detector
  • Other work - statistical machine translation

Arthur Chan AI Advisory (2024-now)

  • independent AI advisors
  • Scientific Advisor of C-Level Managers on the matter of Conversational AI.
  • Two major projects:
  • Lead Researcher in ASR development for European languages
  • Lead Researcher in TTS development for two Asian languages