Usecase
LT Fincorp
- This is a ficticious fin corp. This company is NBFC. This company gives loans to individuals and organizations
- They have a large sales team who interact with customers
- Right guidance to sales team on policies.
- Workflow:
- Sales person will be assigned to a customer by manager
- The initial data about customer is fetched
- Aaadhar Number
- Which leads to
- Good credit history
- Bad Credit history
- No Credit History
- Which leads to
- Aaadhar Number
- Loan Types:
- Personal Loans
- Mortages
- Home loans
- Sales person needs to understand the customer needs
- amount:
- interest rate:
- purpose:
- Sales person enters these details into a system which decides
- intrest rate
- amount (eligible)
- LT Fincorp is losing business because of slow processing
Problem 1: Salesman Guidance
- LT Fincorp has documentation of all the possible scenarios (private)
Comeup with possibilites.
- We need to build a solution which abstracts all the document and policy complications with simple natural language.
- This is RAG.
Problem 2: Quick processing
- Sales representative, takes consent and gets the documents filled by customer
- End of the day they upload the docs to the system
- System is a rule engine with manual approvals/edits
- application => interact with rule engine (decreased)
- Inputs:
- proofs (digi locker)
- requests:
- type of loan:
- amount
- repayments:
- rule engine
- credit score
- income proofs
- give possibilities (rules)
- option 1
- option 2
- option 3
- Agent
LT EdTech
- I train students for competetive exams like NEET, IIT
- Students have lots of different doubts depending on their background.
- I want a patient teacher kind of solution
- understand student
- make student better
- Solution: Fine tuning.
