H.

STARTUP
CONNECT

The operating system for early-stage founder/builder alignment. Powered by AI validation agents.

[ v1.0.4 ][ PROD ]
The Friction

Brilliant ideas die in the "Validation Gap".

Founders often build first and validate later. Builders waste equity on dead-end projects. There was no data-driven bridge between "Gut Feeling" and "Product Market Fit".

“90% of startups fail because they build things nobody wants.”
FIG 01. THE FILTER
Raw Idea Input
AI Validation Layer
Market Report

System Architecture

Orchestrating autonomous agents for real-time due diligence.

01

Input Ingestion

Founders submit unstructured text. NLP parsers extract core value propositions, target verticals, and monetization models immediately.

02

Multi-Model Swarm

Orchestrating 14+ specialized agents, dynamically routing each task to the optimal model:
• Research: OpenAI / ChatGPT
• Engineering: Anthropic Claude
• Specialized: Self-trained Ollama models

03

Compatibility Match

Vector comparison between the validated idea and builder portfolios. Scores imply technical and cultural fit.

The Match Engine

We moved beyond simple tag matching. The core algorithm assigns weighted scores based on:
1. Tech Stack Overlap (Critical)
2. Industry Experience (Bonus)
3. Availability Windows (Filter)

TypeScript Strict Scoring
Postgres Vector Search
discovery.service.ts
1// Calculate compatibility score (0-100)
2async function calculateMatchScore(
3  founderReqs: Reqs, 
4  builderProfile: Profile
5): Promise<number> {
6  let score = 0;
7  
8  // 1. Stack Overlap (Weight: 50%)
9  const overlap = intersection(
10    founderReqs.stack, 
11    builderProfile.skills
12  );
13  score += (overlap.length / founderReqs.stack.length) * 50;
14
15  // 2. Vector Semantic Match (Weight: 30%)
16  const embedding = await openai.createEmbedding(founderReqs.desc);
17  const similarity = cosineSimilarity(
18    embedding, 
19    builderProfile.embedding
20  );
21  score += similarity * 30;
22
23  return Math.round(score);
24}