ARSAM / PROJECTS 05 PROJECTS

// SELECTED PROJECTS

DOCINTEL BACKEND

APIs for document processing and retrieval.

work /
Opened DOCINTEL BACKEND. Backend architecture & AI pipeline development.
OPEN / docintel_backend/SOURCE AVAILABLE
DOCINTEL BACKENDBACKEND API / SOURCE AVAILABLE

BACKEND / AI API

DOCINTEL BACKEND

FastAPI document intelligence API with cited RAG answers, spreadsheet analysis and SSE progress updates.

Public source code is available on GitHub. The backend API is not deployed.

FASTAPI
PYTHON
POSTGRESQL
PGVECTOR
OPENAI
SQLALCHEMY
Source code(opens in a new tab)Backend architecture & AI pipeline development
Project details
My role, technical approach and tools

APIs for document processing and retrieval.

A user-scoped FastAPI backend for document intelligence. It processes PDF, DOCX, TXT, CSV and XLSX files, stores OpenAI embeddings in PostgreSQL with pgvector, and generates summaries and retrieval-grounded answers with source citations. Spreadsheet analytics, downloadable PDF reports and durable SSE processing events support the document workflow.

01 / SYSTEM MAP

  1. APIAuthenticated document servicesFastAPI / JWT / Pydantic
  2. RETRIEVALEmbeddings & cited answersOpenAI / PostgreSQL / pgvector
  3. PROCESSINGAnalysis, reports & progresspandas / ReportLab / SSE

02 / MY CONTRIBUTION

Backend architecture & AI pipeline development

  • Implemented JWT authentication and owner-scoped document, folder and tag APIs.
  • Built upload validation, background text extraction, chunking and vector storage for supported document formats.
  • Connected semantic retrieval to AI summaries and cited Q&A with follow-up question history.
  • Added spreadsheet statistics, cached AI insights and downloadable PDF reports.
  • Implemented database-backed status events and authenticated SSE streams for processing updates.

Published source, not a live API. The current implementation uses FastAPI background tasks and local upload/report storage.