SubMeter Bill: fair bills for a shared meter.
It works out the electricity bill for a tenant on a sub-meter. Enter this month's reading, or snap the meter, and it applies the real tariff: base rate, slab-wise subsidies and the fixed demand charge, preset to UPPCL rates.
Sub-metered homes are everywhere in India, and the monthly bill is usually worked out on a calculator against a tariff sheet. It's slow, easy to get wrong and hard for the tenant to check in detail. I wanted a two-minute bill that both sides can trust.
Loan Saathi: a Hindi voice agent on the borrower's side.
A voice agent can do more than just provide information and handle customer support for companies. It can help understand and analyze requirements clearly and improve users' situations for different scenarios.
Lenders, NBFCs, and distributors often sell the products that earn them the most margin. Borrowers don’t always have someone on their side. Many people in smaller cities are pushed into high-cost debt because of limited financial awareness, and may never revisit those loans.
I focused on people who are more comfortable speaking Hindi than using a chat interface, and who may be stuck in high-interest loans. Since loans are often sold over the phone, a voice interface felt like a natural way to meet people where they are. A failed autopay and bounced EMI acted like a perfect trigger for the user to discover such a product.
Let the LLM talk, not calculate.
Every number came from code. In one early test, the model said 67 instead of 63, and in lending, that kind of mistake can break trust.
More experiments, on the way.
I love experimenting and building things that are cool and useful. New experiments will land here as I build them. Want to build one together? Let's connect for a quick call.