SolarSaathi — AI Solar Bill & Vendor Quote Checker
SolarSaathi uses multimodal LLMs to analyze electricity bills, estimate solar costs, subsidies, savings, and CO₂ reduction, and verify vendor quotes for risks.
Imagine someone comes to your home and tells you that installing solar panels will save you money for the next 20 years.
They recommend a 3 kW system, hand you a quotation full of equipment names and technical terms, and tell you there's a government subsidy too.
Sounds great. But would you know if 3 kW is actually right for your home? Or whether the quotation is reasonable? Or how many years it will take to recover the money you're spending?
We realised something while thinking about this problem as college students: installing solar isn't just an engineering decision. For most families, it's a pretty big financial decision. And the information needed to make it is surprisingly difficult to understand.
That's what led us to build SolarSaathi.
The problem isn't a lack of sunlight. It's a lack of clear answers.
Rooftop solar has a lot going for it. It can reduce the electricity a household needs to buy from the grid, and it can reduce the emissions associated with that electricity. Under the government's PM Surya Ghar scheme , eligible residential installations can receive central financial assistance of up to ₹78,000, depending on capacity and scheme requirements.
But there's a gap between knowing that solar is a good idea in general and knowing whether it's a good idea for your house.
Electricity bills are confusing enough on their own. There are units consumed, fixed charges, different tariff slabs, subsidies and surcharges. A bill from one summer month may look completely different from a winter bill.
Then come the solar quotations: panel brands, inverter capacity, DCR requirements, warranties, additional installation charges and payment schedules. Most people aren't going to spend a week researching every line item. They shouldn't have to.
And one question gets overlooked: what if solar doesn't make financial sense for you right now? A good tool should be able to say that too.
Don't solar calculators already exist?
They do, and they're useful. The government's solar portals and other calculators can estimate system size, cost and savings when you enter the necessary details.
But that still leaves two awkward jobs for the homeowner.
First, you need to find and understand the information on your electricity bill before you can enter it correctly. If you use just one month's consumption, you might miss how much your usage changes through the year.
Second, even after a calculator recommends a system, you still need to decide whether the actual quote from an actual installer looks reasonable. That is a different problem from calculating potential savings.
We wanted to connect those two decisions. Not just "Should I consider solar?" but also "Am I being offered the right system, with the right details, at a reasonable price?"
So we built SolarSaathi.
The idea is simple: take a photo of your electricity bill, understand your options, and then check the quotation before you commit.
Step one: Start with the bill you already have
Instead of asking you to figure out your electricity usage yourself, SolarSaathi lets you upload a photo or PDF of your existing bill.
An AI model reads the document and pulls out details such as electricity consumption, sanctioned load and previous months' usage when they're available. You then get to review and correct those details. That last step matters because a blurry bill or a misread number shouldn't decide how much money someone spends.
Using your confirmed information, location and a set of tariff and solar-generation assumptions, the calculation engine estimates a suitable system size, installation cost, applicable subsidies, potential savings and payback period.
But the point isn't to throw six numbers onto a dashboard. The result should answer questions a homeowner actually has:
How much could my monthly bill go down? How much would I need to pay upfront? And when might I earn that money back through savings?
One of our prototype test bills, for example, has average usage of about 367 units a month. In that test, SolarSaathi estimated a 2.5 kW system with a payback of roughly 5.6 years. Those figures are model outputs using the test assumptions, not a promise that every household with similar usage will get the same result.
We also built the calculation flow to explain where its numbers come from rather than asking people to blindly trust a result.
Step two: Check the quote before saying yes
This is the part we find especially useful.
Suppose you've already decided to explore solar and a vendor sends you a PDF quotation. SolarSaathi can read it and check things people commonly overlook: the price relative to a reference range, panel and inverter details, warranties, included work, advance-payment terms and statements about subsidy eligibility.
A missing detail doesn't automatically mean the installer is dishonest. It means you should know what to ask before paying.
For instance, if the quote mentions a panel brand but not the exact model, SolarSaathi can flag that the eligibility details need clarification and suggest a question you can send to the installer.
We wanted the report to feel less like a technical audit and more like advice from someone who has actually read the paperwork.
There's another kind of saving that doesn't show up on your bill.
Money is the obvious reason many families consider rooftop solar. But it isn't the only reason.
A solar panel generates electricity from sunlight. When that power replaces electricity that would otherwise have come from the grid, it can reduce the need for generation from fossil fuels and the carbon dioxide released in producing that electricity.
Here's a simple way to think about it: if your rooftop generates part of the electricity your home uses, the grid has less electricity to supply for that portion. The exact emissions avoided depend on the grid's generation mix, when the solar power is produced, and the system's actual output.
That's why SolarSaathi also estimates CO₂ emissions avoided each year, alongside financial savings. It uses projected solar generation and an assumed grid-emission factor, rather than inventing an environmental number with AI. India's Central Electricity Authority publishes grid-emissions data that can be used to ground this estimate.
Of course, solar panels aren't magically impact-free. Manufacturing and installing equipment has an environmental footprint too. And an estimated avoided-emissions figure isn't a direct measurement of a household's carbon footprint. Still, the shift toward cleaner electricity matters, especially when adopted across many rooftops.
For us, that's one of the nicest parts of this project: a decision that may help a family save money can also contribute to using less carbon-intensive electricity.
The tech is doing the reading. The maths isn't a guess.
We used React for the interface, Amazon Bedrock for reading bills and quotations, and the Strands Agents SDK in the quote-review workflow. The financial calculations and checking rules are implemented in Python so they can be tested and explained independently of the language model.
We also prepared an AWS serverless deployment with Lambda, API Gateway, S3 and DynamoDB using AWS SAM. At the time of writing, the working demo is running on a local Python server connected to real Amazon Bedrock inference, because our deployment account wasn't verified before the deadline. We'd rather be clear about that than call something "deployed" when it isn't.
We're starting with a limited Delhi-focused setup instead of pretending we've solved every state's electricity tariffs and solar policies. Getting the rules right matters more than having a long list of supported locations.
And we specifically tested the opposite case too: a household with low usage whose electricity bill may already be covered by an applicable subsidy. Sometimes the useful answer isn't "buy solar." It's "check whether it would actually save you enough."
The bigger idea
The more we worked on SolarSaathi, the more obvious the underlying problem became.
People aren't short of solar advertisements. They're short of clear, independent explanations when they are about to spend real money.
We're still improving the accuracy of the policy data, electricity-bill modelling and equipment verification, and this is a decision-support tool—not a replacement for a site survey, official scheme eligibility checks or professional advice.
But if someone can go from staring at a confusing bill and vendor quote to asking the right questions with confidence, that's already a useful step.
Going solar should start with understanding, not guesswork.
If you'd like to see how we're building it, here's our SolarSaathi GitHub repository . I'd genuinely love to hear which part of choosing rooftop solar you find most confusing.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article