Skip to main content

Workshop · Online · Past event

Dating Application Powered by Langsat

Build a compatibility model for a dating app on Langsat — from raw user, match, and message tables to a live prediction endpoint, in one hour.

Date
Sunday, August 30, 2026
Time
Time to be announced
Duration
1 hour
Format
Online · Workshop

What you'll walk away with

Leave with a trained compatibility model running on your own schema, and a REST endpoint your app can call — without writing feature-engineering code or standing up infrastructure.

Why this session

Building the next generation of matchmaking? Langsat is a developer-first platform for training, fine-tuning, and deploying custom dating prediction models.

Whether you are predicting compatibility from user behaviour, shared interests, or conversation activity, Langsat does the heavy lifting so you can focus on the app itself.

Matchmaking is the textbook case for relational machine learning. A hand-tuned compatibility score reduces every user to a flat row of attributes and throws away the thing that actually predicts a match: who already talked to whom, which profiles behave alike, and how interest propagates through the network. Langsat trains directly on those relationships.

Schedule

1 hour, 6 segments. Times are offsets from the start.

  1. 00:00

    What we are building

    The shape of the problem: users, profiles, matches, messages — and what "compatible" means as a training target.

  2. 00:05

    Why matchmaking is link prediction

    Predicting a match is predicting an edge between two users, not scoring one user in isolation. We show what a flat feature table loses.

  3. 00:15

    Connect the data — live

    Point Langsat at a Postgres schema, let it detect tables, keys, and relationships, and review the graph it builds.

  4. 00:30

    Train the model

    Pick the target, train, and read the results — no feature engineering, no notebook, no GPU to provision.

  5. 00:45

    Deploy and call it

    Ship the model to a REST endpoint and score a live pair of users from the app side.

  6. 00:55

    Q&A

    Open floor — bring your schema.

Who it's for

  • Developers and founders building dating, social, or matchmaking products
  • ML engineers tired of maintaining hand-built feature pipelines
  • Anyone with a users / matches / messages schema and no ML team

You'll leave with

  • A compatibility model trained on your own tables, not a toy dataset
  • A live /predict endpoint you can call from your app the same day
  • A clear read on when link prediction beats a hand-tuned scoring formula
  • The recording and the worked schema afterwards

Build this on your own data

Point Langsat at your data, pick the task, and train a real model — no feature engineering, no infrastructure to run. Your data stays private: isolated to your account and encrypted in transit and at rest.