Products built for real-world constraints.

Working products across native apps, accessibility, real-time audio and applied machine learning.

Manualis pipe-organ console with complete iPhone registration and settings screens01
Native iOS / AudioIn development

Manualis

A real-time virtual pipe-organ engine and professional iPhone and iPad console.

Manualis combines a mature sampled-organ engine with an adaptive iPhone and iPad experience. Its real-time path is allocation-free, while the product around it handles sample libraries, MIDI routing, registration, recording and export.

  • Swift
  • C++
  • Core Audio
Marques Talk phrase bank, voice controls and backup settings02
Native AndroidIn development

Marques Talk

Accessible AAC communication for Android with custom input, prediction and voice systems.

Designed for people with speech difficulties, Marques Talk reduces the effort required to construct and speak a message. It combines accessible interaction design with local language data, cloud voice options and robust offline fallback.

  • Kotlin
  • Compose
  • TTS
Two calibrated machine-vision cameras connected to an edge-compute unit for visual speech research03
Machine-learning researchResearch

Multilingual visual speech recognition

Research and prototyping around language identification and speech transcription from silent video.

This research implementation investigates multilingual lip-reading across a large, varied dataset and includes a web prototype for testing real-world video inference.

  • Python
  • PyTorch
  • Transformers
  • FastAPI
  • CUDA

Bring us the whole project or the hardest part.

Tell us what the product or system needs to achieve, where the main risks are, and what has made it difficult.