Privacy Spell

@jonathan_gheller1 This is a fantastic idea. How do you identify, which text pieces should be obfuscated? Does it work with other PII, (e.g., email address, SSN) This product deserves way more upvotes than it currently has.

@salil_sethi The first production version will have the ML model plus heuristics run fully local. Email, SSN, names, numbers...all included. thanks for the good vibes Salil, much appreciated

@salil_sethi @jonathan_gheller1 yup, more specifically, it's a couple of NLP models plus a bunch of heuristics. In our current prototype, part of that is done with a model remotely (of course, sending the information encrypted and securely, and without storing anything anywhere), but the first release we are working on will have the entire model embedded so it will be 100% local detection

@vinnie_t That's great! can you give me an example of the pain this would be solving for you?

@jonathan_gheller1 I work for an education company and so have to follow FERPA guidelines (think HIPAA). That means spending time creating environments with dummy data to fit particular demonstration needs, rather than just using some production data. Obviously the former is the route we should be going, but it also would help protect us in the event a sales person accidentally navigated to something with student/university information.

@jonathan_gheller1 @vinnie_t a family member works at a school district she he was part of the inspiration for this product

BOOYAH!

@jatin_chaudhari Awesome! could you share an example of two of how you would use it? really curious to learn more