Skip to content
texlyBook a call
[ ← WORK / AI ]BEKUR · CASE STUDY · AI PRODUCT + MODELS

The doctor talks.The note writes itself.

Bekur is our own product: clinical documentation for Amharic-speaking clinicians. It shows both halves of what we do: the product people use, and the models underneath it.

PRODUCT
Bekur, built by Texly
DOMAIN
Clinical documentation
LANGUAGES
Amharic + English
STATUS
MVP, models in training

[ THE PROBLEM ]

Consultations happen in Amharic, with English medical terms mixed in. Notes are written in English, after the fact. And general speech models weren’t good enough at either.

[ WHAT WE BUILT ]

A product doctors record into, and our own models behind it, trained for clinical Amharic speech, translation and summarization.

[ 01 · THE PRODUCT ]

Record, review, done.

HOW WE BUILD AI PRODUCTS →
  1. 01 RECORD

    Record

    The doctor records; audio uploads in short chunks, so sessions can pause and resume.

  2. 02 TRANSCRIBE

    Transcribe

    Speech to text, with who said what and when.

  3. 03 TRANSLATE

    Translate

    Amharic into clear clinical English.

  4. 04 SUMMARIZE

    Summarize

    A SOAP note the doctor reviews, edits or regenerates.

  • Speaker identification

    Voices are linked to doctor or patient, so the transcript shows who said what.

  • Templates per specialty

    Each specialty decides what the note extracts and can adjust it.

  • AI review

    Gaps and suggestions are flagged inside the note for the doctor.

  • Roles & billing

    Doctor, admin and accountant views, with usage and billing at the organization level.

HOW WE TRAINED BEKUR, STEP 01 / 06 · COLLECT

Real consultations, transcribed

We collected clinical consultations and everyday conversation across Amharic dialects, and had them carefully transcribed, including the English medical terms people actually use.

SCROLL FOR THE NEXT STEP ↓

[ WHERE IT STANDS ]

DEMO-READY FOR CLINICS
MVP
CUSTOM SPEECH & TRANSLATION MODELS
Training
VISIT BEKUR.CO ↗
[ NEXT CASE STUDY ]Ledger
Read it