What the weather actually did at this address, for insurers settling claims
TL;DR Exposa answers the question insurers love to argue about - how much rain, wind or hail hit this exact address on this date? It merges station, radar and satellite data into a dated report for one location. Insurers and licensed loss assessors buy it per report or on an annual plan.
Weather claims in India get settled on thin evidence. The nearest official rain gauge can be 20 km from the damaged building, and the storm that flooded one colony may have missed it completely. So claim files rest on photographs, a neighbour's account and a district rainfall figure that averages the event away.
You can see the cost of that vagueness in under-insurance. In a household survey across four coastal Odisha villages, insurance covered about 2% of losses from the 1999 super cyclone and about 4% from Cyclone Phailin in 2013, leaving 77% and 83% of losses uncompensated[1]. India also has long state-level flood damage records going back to 1953[2], but nothing at the scale of a single plot.
What it does, in regulatory terms: it sells weather evidence, not a statutory loss assessment. Survey and loss assessment work is reserved for IRDAI-licensed surveyors and loss assessors under their own regulations[3]. Exposa supplies those surveyors and the insurer with the meteorological facts. It does not value the loss or decide the claim. Confirm the current regulation before you sell anything.
A user enters an address or coordinates and a date range. Exposa rebuilds local conditions from the nearest station readings, radar reflectivity, satellite rainfall estimates and terrain, and returns a report with hourly rainfall and wind at that point, the probability a hail cell passed over it, how the event compares with the past 20 years and a plain statement of how certain each figure is. If the data can't support a conclusion then the report says so.
The seed proposed hyper-local climate modelling as a platform. On its own that has no buyer in India, because modelling is a means and not a product. An insurer paying claims has a dated, specific question and a reason to pay for the answer.
The users are claims managers at general insurers and the licensed surveyors working for them. The payer is the insurer's claims function, or the surveyor firm billing it through. Reinsurers and brokers are a second buyer for event footprints after a cyclone. There's no public price for this in India, so ₹1,500–6,000 a report, or an annual plan for a claims team, is our assumption.
Three things line up. Cyclone and flood losses in India are documented well enough to argue about, with household studies now putting figures on residual loss[1]. Machine learning work on disaster loss and claims estimation has moved from papers to pipelines, including one 2026 study built on 68,485 disaster records across 10 disaster types[4]. And parametric products that pay on measured weather already run at scale, which trains the market to accept a measurement as the trigger[5].
No published figure covers spending on weather evidence for Indian claims. The narrow, honest signal is the gap itself - 2% to 4% of coastal household cyclone losses were insured in the two events studied, with roughly four-fifths uncompensated[1]. The adjacent number is one Indian parametric-insurance tech firm, which says it has built parametric portfolios of over US$5 billion (about ₹48,000 Cr) across 30-plus countries and runs a climate sensor network of more than 15,000 nodes[5]. Treat that as proof the category gets funded, not as your market.
That firm sells parametric platforms and climate risk services, with its own sensor network[5], but it's built around designing and paying parametric cover rather than evidence for conventional claims. Global catastrophe-modelling firms supply catastrophe models and data to insurers and reinsurers worldwide[6], priced and scoped for portfolio modelling. India's own weather agency publishes forecasts and station data for the public[7], not claim-ready reconstructions. The gap is a per-address, per-date report a claims file can carry.
| Difficulty | Moderate (42/100) |
| Starting capital | ₹90 L–1.8 Cr |
| Time to launch | 9–12 months |
Pulling the data together is routine engineering. Two things are not. The first is honesty about uncertainty, because a report that overstates confidence will get torn apart the first time a claim is disputed, so the error bands have to hold up. The second is the sale. Claims teams are conservative, and you'll need a surveyor or insurer willing to run your report alongside their usual process for several months.
You don't need an insurance licence if you only sell weather evidence. Stay clear of the work that's reserved for licensed people.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't let AI decide a claim or value a loss. That stays with the licensed surveyor and the insurer, and where data is thin the report should say so.
Want to work out the AI layer for Exposa? Talk to Yash →
I think it's worth a look if you know meteorology or geospatial data and can bring in someone from general insurance claims. If your team wants to sell forecasts then it's a poor fit. Prove one thing first - take 50 settled claims from one insurer, rebuild each location and show your reports would have changed the outcome or settlement time on enough of them to pay for themselves.
Plan for roughly ₹90 L–1.8 Cr, mostly on data, meteorology talent and a long pilot with one insurer.
Not if you only supply evidence. Loss assessment is reserved for IRDAI-licensed surveyors, so get a lawyer to confirm you're outside the intermediary rules.
Yes, weather claims get disputed everywhere. Check your local insurance regulator's rules on loss adjusting and the licence terms of your weather data.
The free Exposa validation kit has the line-by-line capital breakdown, the effort scoring, the insurance and data-protection checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for claims teams and surveyors, and 5 assumptions to test. Download the kit →
Thinking about building Exposa or something close to it? The make-or-break work is proving your reports hold up when a disputed claim is argued line by line, and that's the plan I'd help you build before you raise. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.