How Delphi Is Building The Social Graph Behind Its Next Growth Loop
Applied AI engineer Kaan Karakas is mapping how people connect around Delphi’s Digital Minds, including people who haven’t joined yet.

When someone starts using Delphi, a platform for creating AI versions of real people, they can build a Digital Mind that others can chat with or call. Some of the people they follow on social media may already be there, with Digital Minds of their own. Those connections give Delphi a way to find people who might have a reason to join before they sign up.
Kaan Karakas, an applied AI engineer at Delphi, owns the company’s newer networking and indexing project, which is key to mapping the relationships around those people, including connections to people who haven’t joined.
While his work is still in its early stages, Karakas wants to use that map to help Delphi find prospective owners and make it easier for existing users to discover someone whose knowledge interests them. Both depend on understanding how people connect beyond the platform’s account list.
Indexing The People Around Digital Minds
Delphi is already creating Digital Minds belonging to public figures, business owners, CEOs, and venture capitalists, Karakas says. But the people who might make the platform useful to a wider audience won’t all be found among its current accounts. “We believe Delphi should become a platform where anyone can create a Delphi and have an AI represent them,” he says.
His networking project starts with the people already connected to those owners. Someone followed by several of them might have knowledge that interests people on Delphi, even if that person has never heard of the platform. To recognize that possibility, the team needs a way to include people in its map before they sign up.
Karakas’s project uses a social graph to do that. Each person is represented by a node, and a follow becomes an edge connecting two people. The graph keeps track of which people have a Digital Mind and which haven’t joined, so a connection to someone outside Delphi doesn’t disappear simply because they lack an account.
That distinction lets the team use the map for two different purposes. An existing owner has a Digital Mind that someone can consult today. A person outside the platform may be worth inviting because people already there have shown an interest in them.
Giving People A Place Before They Join
For someone who hasn’t joined Delphi, a public social handle can give the index a place to start. The project includes plans for stub profiles: records that hold a person’s connections before they claim an account. Follows and contacts can then be attached to that record instead of remaining separate links with no shared home.
If several Delphi owners connect to the same handle, those relationships can be gathered around one person. Each additional connection helps the team see how that person fits into the network. The record becomes useful before its owner does anything on Delphi because it shows who already has a connection to them.
Karakas sees that social activity as a useful clue to whose knowledge people might want to explore. “If a person follows or is followed by someone on a social media app, they might also be interested in that person’s Delphi,” he says.
A follow doesn’t establish that either person wants to use Delphi. But gathering those links into a stub profile gives the team something more specific to consider: an outside person connected to several existing owners, rather than an unfamiliar handle in a single account’s contact list.
Using Existing Connections To Guide Invitations
Karakas says Delphi has found substantial overlap between its owners and people who haven’t joined. Some outside people connect to several existing Delphis, giving the team a reason to consider what they might gain from an account of their own.
He offers a hypothetical example: “Let’s say someone isn’t on the platform, but they have 60 Delphis connected to them. That means they would most likely find value in the Delphi platform.”
The potential benefit runs in both directions. That person would arrive with connections to people already represented on Delphi. Their own Digital Mind could also give existing users another person to consult. Recording more of those relationships helps the team see which prospective owners have ties to the network it has already built.
Warm invitations are one way Delphi could use that information. The graph helps identify people worth approaching, while the invitation gives them a chance to decide whether joining would be useful. A densely connected record is a starting point for outreach, not a substitute for that decision.
Karakas says the project is adding five- to six-figure numbers of edges, giving the team a larger map of the relationships between Delphi owners and people outside the platform.
Finding Someone Who Can Help With A Question
For Karakas, bringing more people onto Delphi matters because of the knowledge they could make available. He describes an interest in robotics and imagines wanting to learn about work at a leading robotics lab . Reaching a researcher directly might be difficult; finding their Digital Mind could offer another way to explore the subject.
“If I could find a robotics researcher on the Delphi platform, I would essentially have an open book. I could ask any question I had,” he says.
In that scenario, the researcher wouldn’t need to reply personally to every inquiry. Someone could ask questions of the knowledge available through their Digital Mind, even if they had no existing relationship with the researcher.
Karakas’s indexing project addresses the discovery problem behind that encounter. Delphi needs to know which people it already has and how they connect to the wider world. Those relationships can help connect a user’s interest with a relevant person, making expertise easier to find within a growing network.
That also gives the invitation work a purpose beyond adding accounts. A useful new Digital Mind could help someone already on Delphi learn about a subject they couldn’t easily explore before. Better access to that knowledge, in turn, gives other people a reason to consider joining.
Giving People A Reason To Join
Someone who appears in the Delphi graph before signing up may eventually contribute knowledge that makes Delphi more useful to others. Their existing relationships help the team understand where they could fit, while their Digital Mind could give people a way to learn from them once they join.
Kaan Karakas’s indexing work at Delphi helps the team find people through their existing connections, decide whom to invite, and make their knowledge easier for others to discover once they join.
BDG Media newsroom and editorial staff were not involved in the creation of this content.