Back to portfolio

Case study · Northeastern project · Sep 2025 – present

Translation Exchange

Live speech-to-speech translation and cultural-sensitivity insights for a Maine credit union — a Northeastern project, Sep 2025 to present.

Role
Lead engineer on the real-time speech translation pipeline, emotion-aware analysis, and Google Cloud deployment
Building
Exchange
Year
Sep 2025 – present

Problem

A Maine credit union needs live translation and communication analysis that surfaces cultural-sensitivity insights across its operations and branches.

Daniel's role

Built an LLM-powered translation and communication-analysis product spanning a speech pipeline, call UI, and Cloud Run deployment for six branches covering $415M in operations.

System diagram

Live translation and cultural-sensitivity analysisLive LLMCulturalsensitivityCloudRun

What I built

  • Speech-to-speech translation pipeline with emotion detection from tone of voice (Hume)
  • Automatic form filling during calls (mood, name, phone, notes)
  • Streaming transcription for live use
  • Text-to-speech routing for African accents
  • Full UI rebuild with a one-button call interface
  • Move to Cloud SQL, Cloud Storage, Redis, and Cloud Run deployment

Stack

  • React
  • TypeScript
  • Node/Express
  • OpenAI
  • Google Cloud Speech/Translate
  • Faster Whisper
  • Postgres
  • Cloud Run

Outcome

Deployed on Google Cloud Run with cultural-sensitivity insights across the credit union's $415M in operations and six branches.

Field notes (measurements)

  • Operations

    $415M

  • Branches

    6

  • Deploy

    Google Cloud Run

Find this building on the city plan →