Conversational AI engineer working across large language models, speech synthesis, and cognitively-grounded language interfaces. Currently building Flowent, a real-time voice agent for language immersion.
Selected work
Production systems and research spanning conversational AI, speech synthesis, and audio.
A production voice agent: streaming bidirectional audio, sub-second turn latency, and multi-turn speech interaction, grounded in cognitive-science immersion methodology.
Live in production on the App Store and Google Play, with sub-second turn-taking and streaming TTS.
A text-to-speech system that conditions prosody and emotional expression on visual cues, producing more natural, context-aware synthesized speech.
An AI-driven workflow that generates high-quality TTS training datasets, removing the data bottleneck for expressive speech models.
On-device text-to-speech for React Native — local inference, no server round-trip, private by default.
Notes & essays
Technical posts on speech, language models, and the things I build.
We built on-device neural text-to-speech for Flowent because React Native had no good answer — 31 languages, no API keys, no network calls, no per-request bill. Now it is open source. Here is the path from install to shipping, rough edges included.
Flowent — the real-time voice agent for language immersion — is now available to download on iOS and Android.
Two vision-language models, the same 150 silent RAVDESS clips, one question — can a VLM read emotion from a face with the audio stripped out? A head-to-head on classification accuracy and per-class behaviour.
What I work with
The stack I reach for when building speech-first, language-driven systems.
Try it yourself
Interactive models and apps you can run right now in the browser.
A production voice agent: streaming bidirectional audio, sub-second turn latency, and multi-turn speech interaction, grounded in cognitive-science immersion methodology.
A GPT-2 language model implemented from scratch in PyTorch, trained to generate Shakespeare-style text.
Real-time music style transfer using RAVE (Realtime Audio Variational autoEncoder) for creative audio manipulation.
On-device text-to-speech for React Native — local inference, no server round-trip, private by default.
Let's talk
I'm open to consulting, collaborations, and interesting full-time roles in conversational AI and speech. The fastest way to reach me is email.
