Solutions
Graph-native visual entity resolution, on your infrastructure.
PhotoGraph reads an entire photo collection, decides which images show the same person, and stores the answer as an identity graph you can search, query and build on. It needs no labels and no training, and it runs inside your security perimeter.
At a glance
Solution summary
- What it resolves
- People across a lifetime
- Input
- Scanned prints, slides, negatives, contact sheets, document scans and born-digital photos
- Labels or training
- None required
- Core technology
- Graph Resolution Core (GRC), patent pending
- Output
- Queryable identity graph
- Interfaces
- We integrate with your UI and APIs, or we can help if you need them
- Deployment
- Your environment, your data
- Data egress
- Usage statistics only. Images, results and metadata stay in your environment
- Scale
- Architecture designed for billion+ edge identity graphs. We scale with you
- Customize
- We can provide one solution or several. Have your own face recognition algorithm – we’ll customize PhotoGraph to work with it. Don’t have one – we’ll help identify and install the right one for you.
Architecture
Everything happens inside your environment.
Photos are read where they already live. Resolution, the identity graph and every interface run on your hardware.
Graph Resolution Core
Where facial recognition stops, we keep going.
Our patent-pending Graph Resolution Core (GRC) holds one person together as one identity across decades, and doesn’t confuse look-alike relatives. A face that is partly hidden, or not visible at all, is not the end of the search. It is one more photo of the same person.
Every appearance, not only clear faces
Group shots, people half-hidden behind others, backlit subjects and photos taken from behind are resolved alongside the portraits.
Aging is expected
A toddler on a 1950s slide and the same woman at 76 belong to one identity. Decades of change are part of the model, not an error.
Relatives stay separate
Siblings, parents and children who look alike are kept as distinct people instead of being merged into one.
Works with your entity resolution
The perfect complement to text-based entity resolution.
Most entity resolution matches records on names, IDs, addresses and dates. PhotoGraph uses pixels only, so you can now find identities independent of the textual data, and add a second, independent line of evidence to the systems you already run.
- Built for fraud. Perpetrators intentionally use entirely different keys, names and historical data. The text never lines up. Their photos still connect.
- Independent evidence. Every photo link stands on its own, with no text field behind it.
- Integrates with what you have. We can customize outputs to integrate seamlessly with your existing entity resolution dataset, or consult with your team on integration methods.
Illustrative example. Names and records are fictional.
Identity graph & API
Results you can query, not just look at.
Every resolution is stored as a structured identity graph: people, the photos they appear in, who they appear with, and when. Your teams, tools and AI systems read it directly.
- Graph query API with endpoints ready for downstream integration.
- Feed your catalog or product: push person links into a collections system, a subscriber search feature or a case tool.
- Built for scale: an architecture designed for billion+ edge identity graphs. We scale with you.
What the graph answers
- Every photo of this person
- Who appears most often with my target or source image
- Which donor collections include the same people
- Photos where this person appears but no one is named
- The Kevin Bacon game: find the fewest connections between two or more people
Plain-English questions
Ask the archive, in plain English.
An optional language-model interface uses graph-augmented generation, so non-technical staff can ask who appears where across an entire archive. Answers are grounded in the identity graph, and the language model can run on your own hardware too.
Example questions
“Show every photo of these two people in the same image.”
“Who appears with him in the fire department photos?”
“Which donors gave us photos of the same family?”
Deployment
Sovereign by design.
PhotoGraph runs where your collection lives. No images, results or metadata ever leave your environment. The only thing reported back to us is usage statistics.
On-premises
Full operation inside your environment with no cloud dependency.
Your hardware
Runs on hardware you already own. We size the deployment to your collection during the pilot.
Works on day one
No data collection, labeling or model training before you see results.
See the solution on your photos.
A demo walks through real results. A pilot shows them on part of your own collection.