SunderStar — Geolocate any photo, anywhere on Earth

How it works

We read the landscape, not the street signs.

Extract the terrain signature

From a single photo we isolate the geometry the landscape imposes — skyline profile, ridgeline contours, and the edges of coast, lake, and river.

Match against global elevation

That signature is searched against satellite imagery and digital elevation models spanning the entire planet, narrowing millions of candidates to a handful.

Return coordinates and confidence

You get camera position, viewing direction, and a calibrated confidence score — evidence you can show your work with, not just a pin on a map.

Every stage stays editable — the analyst corrects the region, bypasses a filter, and re-runs from cache. Ruled-out branches stay on the canvas as evidence.

Operator control

Control how your agent investigates.

An agent that can only hand you an answer is one you have to take on faith. Every stage of the search is a node you can halt, correct, or switch off — mid-run, without starting over.

Stop it and tell it what you know

The search commits to the wrong region and starts spending on it. The analyst halts the run, replies to that step — the post was tagged in Alberta — and it re-ranks and carries on from there. The branch it ruled out stays on the canvas as evidence.

Switch off a filter you don't trust

A land-cover layer from 2019 prunes away terrain that burned in 2023, and the match comes back empty. The analyst bypasses that one filter — not the whole run — and only the downstream stage recomputes.

Why we're building it

What if every photo could prove where it came from?

Skilled analysts already do this by hand — cross-referencing ridgelines against maps for hours per image. It works, and it does not scale to a breaking news cycle or an active search. We think that capability should belong to everyone who needs it.

Hours → seconds

Manual terrain analysis, automated end to end

100%

Of Earth's landmass covered by elevation data

1 photo

No metadata, no EXIF, no Street View required

Team

Built by people who needed this to exist.

Mohammad Jihad

Co-founder

Studies computer science at Harvard and builds open-source tools for exoplanet discovery — finding faint signals buried in noisy sky data.

1:1

Co-founder

Actively interviewing

We're looking for a second founder with deep operational experience — someone who has worked in the field where locating a photo is urgent, not academic. Get in touch.

Get in touch

Tell us what you need to locate.

We're onboarding early users now. If you work in verification, response, or rescue, we'd like to hear about your workflow.

hello@sunderstar.tech Read the method