Skip to content

Proposalreview

Problem:

Thousands of proposals received each month for federal funding opportunities.

  • Need to make decisions on whether or not to fund companies quickly.
  • Millions of dollars spent on which companies to fund, and how to fund them.
  • Want to fund the best companies with the most robust AI enabling technology.

Process looks like: - government identifies a need and puts out a soliciation. Ideally this solicitation is very well thought out. - government advertises solicitation to people via emails, marketing, newsletter - government receive many proposals for the solicitation, and needs to review them. - government initially want to filter out bad proposals. - government route proposals to the right person to review - ideally subject matter experts - government emails each reviewer them the link / documents to review - each reviewer goes and reads the document entirely by hand, assigns a score based on the rubric - admin merges all the predictions / reviews to get a final score - ideally have an AI model help speed up the review process by extracting info from each proposal and giving a guided score - the AI model needs to be accurate (answers + sources) and explainable, that way AI assisted info is relevant and tailored. - input the results from the proposal evaluation into valideval for each item. - government does final compliance check, and then awards people the funding based on the proposal they do.

Classification with Anote

Categories include:

-

-

  • Other: (programmatic labeling for this category)

Emailing Reviewers with Upreach

Joining the table with people and categories from the download page - For each person, create the link and send them an email with upreach

Prompting and Merging with Anote

- They log into Anote, want to look at the proposals to tag