Skip to content
Start a project
Start a project

[email protected]LinkedIn

Client projectClient confidential

Recorded calls and scanned documents in, clear structured summaries out.

Services
AI Engineering
Client
Confidential
Built with
Python · AssemblyAI · Gemini API · Tesseract OCR · Flask · SQLite · React
The landing page: every conversation and every document, clear in seconds, with call and document summaries side by side.

01

Important information was spread across call recordings and paper documents, and reviewing it all by hand took too long.

02

  1. 01One summarisation pipeline for both audio and documents.
  2. 02Call transcription with speaker separation, so each voice in a conversation is identified.
  3. 03Text recognition (OCR) that reads scanned and printed pages.
  4. 04A shared workflow that summarises calls and documents in the same structured format.
  5. 05A Flask backend with SQLite storage.
Built with
  • Python
  • AssemblyAI
  • Gemini API
  • Tesseract OCR
  • Poppler
  • Flask
  • Jinja2
  • SQLAlchemy
  • SQLite
  • Flask-Login
  • Flask-WTF
  • WTForms
  • Gmail SMTP
  • React
  • Vite
  • Bootstrap 5
  • HTML
  • CSS
  • JavaScript

The product

The dashboard: upload a call recording or a document, then read the transcript by speaker, the AI summary, key points and action items.
Audio calls: recorded and uploaded calls with a speaker-labelled transcript, key points and action items.
Documents: scanned and uploaded files read with OCR, each with its extracted text, an AI summary and key topics.
History: every call and document summary, with its status and details.

Representative interface; the client's product is confidential.

Next project

Client confidential

Medical Knowledge Chatbot

Read the case study