KinkLink

Matchmaking service

  1. User registration Users create profiles that are managed by their individual agents. Each agent holds the user’s preferences, profile information, and interaction history. Agents use Zupass for verifiable credentials, ensuring that users are real and preventing catfishing.

  2. Matchmaking Process: The multi-agent system runs the matchmaking algorithm. This system only needs pseudonymous information (e.g., agent IDs and Elo ratings) to function. Agents periodically request matchmaking updates from the multi-agent service. These updates include potential matches based on the algorithm.

  3. P2P AI-Driven Interactions: Once a match is made, the individual agents use Waku to establish a secure and private communication channel. Conversations are designed to mimic human interaction, allowing agents to gauge compatibility and engagement.

  4. Feedback Mechanism: After an interaction, agents provide feedback to the multi-agent service about the experience. The agent summarizes this feedback into a non-identifiable success metric (e.g., interaction success, engagement level) and sends it to the multi-agent matchmaking service.

  5. Updating Matchmaking Algorithm: The multi-agent system uses the received feedback to update the matchmaking algorithm (e.g., adjusting Elo ratings). This feedback loop helps the system improve match quality over time while preserving user privacy.

Team's submissions

KinkLink Infrastructure

The problem KinkLink solves

Traditional dating apps require users to constantly swipe and make decisions about potential matches, leading to decision fatigue and cognitive overload. Moreover, many people cannot express themselves freely, out of fear of prosecution or worse. Even in more liberal places, users may feel hesitant to express their true preferences and identity on dating platforms due to concerns about privacy and judgment. And lastly, existing dating platforms often rely on proprietary algorithms for matchmaking, leaving users in the dark about how matches are made.

Challenges you ran into

We could not get the Waku demo repo to work reliably and decided to leverage libp2p instead. We also ran into the issue of dependency hell in python. It took a nightmarish amount of time to get the FSM application configured properly with compatible part and a considerable amount of hacking and a bit of monkey patching was required.

Technology used

Autonolas, Tendermint, HOPR RPCh, Libp2p, Gnosis Safe, Docker, Solidity