Dexter

Building Dexter, a secure and searchable knowledge for consultants. From our consulting experience, we saw that consultants cannot easily share or upload slide information from cases because of confidentiality concerns (i.e. slides contain the client name and revenue financials). As a result, it’s hard for people to store information in an internal database and query it to learn more information about different case work. Our solution is a web platform that lets the user upload slides and uses GPT-5 to apply redaction rules (i.e. filtering out client names, financials, and rewriting information to make it less identifiable). We use C# to parse the underlying XML of the slides, then pass the information to the python layer for the AI processing step. Specifically, we use GPT-5 to analyze the text as well as images. Then we store the information in Supabase, which we access later when the user makes a query.