DIANNT
WorkPlatformsEventMatch SF

Platforms / 2026

EventMatch SF

An event-ingestion and matching system that turns personal interests into useful San Francisco recommendations.

A public TypeScript platform joining event scraping, preference onboarding, semantic search, and conversational recommendations.

EventMatch SF preference screen with selectable Friday night activities
Concrete social scenarios make preference capture feel human instead of taxonomic.
Role
TypeScript product engineer
Timeframe
2026
System
Platforms
  • TypeScript
  • Next.js
  • semantic search
  • pgvector
  • scraping
  • recommendation systems
01

Overview

The system in context.

EventMatch SF collects local event data and builds a preference profile through a guided questionnaire and optional social-link parsing. The resulting signals feed a matching model designed to explain why an event fits, not simply return a generic feed.

The architecture described in the public repository spans Eventbrite ingestion, scheduled refreshes, inferred tags, a pgvector-ready schema, hybrid search, and both web and Telegram conversation surfaces.

What shipped

  • Uses a guided interest questionnaire and social-link parsing to build a preference profile.
  • Documents a six-dimension scoring model with a 100-point result.
  • Combines scheduled event ingestion, embeddings, and hybrid search.
  • Exposes recommendations through both a web experience and conversational interfaces.
02

Measured context

The numbers, with their meaning intact.

onboarding questions
7

Questionnaire length documented by the public project README.

Evidence source ↗ (opens in a new tab)
point match score
100

The documented recommendation score spans six matching dimensions.

Evidence source ↗ (opens in a new tab)
ingestion cadence
6h

Scheduled event refresh interval documented in the repository architecture.

Evidence source ↗ (opens in a new tab)
03

Product record

Screenshots from the product and project record. Open any frame for a closer view.

01Concrete social scenarios make preference capture feel human instead of taxonomic.
02A measured onboarding sequence establishes context before recommendations appear.
04

Build story

From signal to shipped system.

  1. 01 / Ingest

    Build a living event layer

    Scheduled scraping and inferred tags keep the local event corpus structured and current.

  2. 02 / Understand

    Turn preferences into signals

    Questionnaire answers and optional social context become an interest profile rather than a flat category list.

  3. 03 / Match

    Explain the recommendation

    Scoring, semantic retrieval, and chat surfaces connect a person to events with visible reasons for the fit.

Project links

Open the live product, working demo, publication, or repository.

Related work

All work →