CDAO Labeling Orchestration

A cutting-edge custom application designed to empower data engineers to quickly filter, select, and send images for labeling using rich metadata.

This proof-of-concept sets the stage for a fully scalable platform that external data labeling vendors can seamlessly integrate with—speeding up AI training and accelerating insights.

Role: Lead designer

Squad Composition:
Lead Designer
Jr. Designer
Business Technical Leader
Solution Architect
AI Engineers
Front End Developer
Platform Engineer

Challenge Background

Problem: Military systems generate massive volumes of imagery and manually sorting them would be too slow and inefficient.

Solution: An orchestration platform that can integrate with external labeling vendors to reduce delays while allowing the DoD to quickly and securely find, filter, and send only the most mission-relevant images for labeling.

Impact: This speeds AI model updates, improves accuracy, and ensures mission decisions are based on trusted data.

High Fidelity Protoype

Screen recording of the protoype

Dynamic Filtering

Given the complexity of the data and project, I implemented dynamic filtering, where the filter conditions are not fixed but can change based on variable conditions the user inputs.

This enables the data engineer to easily filter metadata and quickly send files for labelling or relabelling.

Design Tradeoffs

Shopping Cart

We used a shopping cart–style pattern for selecting images based on metadata filters.

Due to technical limits, selections couldn’t carry over across multiple image sets, so users had to submit or clear their cart before starting a new search.

While this wasn’t an ideal experience, I included a modal warning that let the user know that if you abandon the cart all selected images would be cleared..