We're looking an ML engineer that can help us create an algorithm that will assist us in background checks.
The process we're trying to automate is part of our due-diligence KYC process. Currently when onboarding new client, part of our KYC process involves searching for adverse media on google. To do this we search the prospective-clients name + several red flag term ('scam', 'fraud', 'money-laundering, etc.).
For example: Satish Bakhda + 'crime', Satish Bakhda + 'conman', Satish Bakhda + 'money laundering'.
This process, however, returns a lot of articles which we then have to go through one by one to confirm whether or not its the same person and whether or not the article is indicative of the person being involved with the red flag term or not.
The software will pull the name from our asp.net portal and use + Scam, + Crime, + charged, etc and returns a shortlist of articles where the prospective-client has potentially been involved in the red flag terms.
Here's what I imagine the finished process to look like:
1. We input features such as name, nationality.
2. Script can use either web scraper or googles API to return a list of articles that may contain adverse news about the client.
3. Classifier algorithm returns percent chance that each article is a positive hit
4. We get a shortlist of articles for human screening based on a threshold for positive hits.
However, if you think that there may be a better approach I'm happy to hear suggestions.
Costs incurred such as the creation of a training set (through something like MTurk) or GPU time (through AWS or Google services) would be paid for by the company.
About the recuiterMember since Mar 14, 2020 Meetul Jain
from Karnataka, India