AI-powered forensic journalism tool. Born in a hackathon.
During an engineering hackathon, I led the design for a new AI-powered reporting tool called Haystacker. We prototyped a product for forensic journalists at The Washington Post that would empower the newsroom to easily find the needle in the proverbial haystack.
By using AI to label to large volumes of media, Haystacker would save the newsroom hours of painstaking labor spent manually labelling each video one by one. This is time better spent on the thoughtful analysis that only the Post's skilled forensic journalists could produce.
Contributions
User interviews
Product vision
Concept ideation
Prototyping
Usability testing
Iterative design
Documentation

The asset-first view empowers journalists to go in-depth with an individual video and its labeled instances.

Haystacker needs to support navigation between an arbitrary number of user-created projects.
From vision to reality.
Due to our success at the hackathon, the team earned the resources to build a usable product. I defined a vision for a maximally useful initial release while laying the groundwork for future features. The first version was asset-focused, with AI processing videos and labeling them for analysis by the newsroom.
The first story reported using Haystacker was published soon after release. 745 campaign advertisements were processed by Haystacker, assisting further analysis by our journalists. I'm proud to have my name included alongside the in-depth journalism facilitated by Haystacker.
This important work continues.

Flow for creating a new Haystacker project from the sidebar.

Feedback flow when uploading media into a Haystacker project.
Want to learn more? Let's talk!
👋🏻
