# Data Collection

Portfolio record by Eugene Livschitz

Canonical page: https://livytech.space/projects/data-collection/

Labeled voice dataset operations

Internal full-stack platform for collecting, reviewing, enriching, moderating, curating, and exporting labeled voice samples for AI, analytics, and research workflows.

## My contribution

Built reviewer and administrator workflows across the React interface, Node.js API, MongoDB data operations, S3 media storage, reporting, bulk ingestion, and AI-assisted enrichment.

## The hard part

Supporting practical high-volume dataset operations while preserving human moderation, traceable review states, searchable metadata, and label-quality controls.

## What shipped

- Built search, editing, filtering, moderation, reporting, exports, audio upload, metadata editing, trait labeling, and pending/good-sample curation.
- Implemented contributor/admin operations, user lifecycle controls, activity statistics, quality flags, and review-state transitions.
- Added CSV-to-file bulk ingestion and AI-assisted trait, biography, and financial-risk suggestions while keeping human validation central.

## Technical details

- JWT-protected reviewer/admin access separates contribution and moderation responsibilities.
- S3-backed audio storage and MongoDB metadata operations support searchable collections of labeled media records.
- CSV import/export and filename matching make batch operations possible instead of requiring one-at-a-time manual handling.

## Technology

React, Redux Toolkit, React Query, Ant Design, Node.js, Express, MongoDB, JWT, S3, CSV, OpenAI