> ## Documentation Index
> Fetch the complete documentation index at: https://withseismic.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Contra: Scaling Marketplace Inventory via Distributed Intelligence

> How we engineered a distributed scraping architecture using 30,000+ Chrome extension nodes to find 5M+ opportunities.

<script type="application/ld+json">
  {`{
    "@context": "https://schema.org",
    "@type": "Article",
    "headline": "Scaling Contra's Marketplace with Distributed Intelligence",
    "description": "Engineering case study: Finding 5 million opportunities using a distributed Chrome extension architecture.",
    "url": "https://withseismic.com/case-studies/contra-linkedin-automation",
    "author": {
      "@type": "Organization",
      "name": "WithSeismic"
    },
    "publisher": {
      "@type": "Organization",
      "name": "WithSeismic",
      "logo": {
        "@type": "ImageObject",
        "url": "https://withseismic.com/logo/light.svg"
      }
    }
    }`}
</script>

<div className="grid grid-cols-1 md:grid-cols-4 gap-4 mb-12 not-prose">
  <Card title="5.7M+ Opportunities" icon="arrow-trend-up">
    Found for freelancers, driving 10x inventory growth
  </Card>

  <Card title="30k+ Nodes" icon="network-wired">
    Active distributed scraping nodes
  </Card>

  <Card title="#1 Product" icon="trophy">
    Product Hunt Product of the Week
  </Card>

  <Card title="Zero Ops" icon="server">
    Serverless architecture with no scraping infra
  </Card>
</div>

## The Inventory Bottleneck

Contra, a commission-free freelance marketplace, faced a classic chicken-and-egg problem: liquidity. To attract freelancers, they needed a massive volume of high-quality job opportunities. To attract clients, they needed active freelancers.

Traditional scraping approaches were non-viable:

1. **Platform Defenses**: LinkedIn and X (Twitter) aggressively block data center IPs.
2. **Cost**: Maintaining a proxy rotation infrastructure for millions of pages is prohibitively expensive.
3. **Context**: Generic scrapers lack the "social graph" context required to find relevant opportunities.

We needed a way to ingest opportunities at scale without triggering anti-bot defenses or incurring massive infrastructure costs.

## Solution: Distributed Intelligence

We inverted the scraping model. Instead of a centralized server farm, we leveraged the users themselves. We built **Indy.ai**, a Chrome extension that acts as a distributed edge node for opportunity discovery.

### System Architecture

The system relies on a "parasitic" architecture (in the symbiotic sense) where the extension piggybacks on the user's authenticated session to read content they are already viewing or that is available to them.

```mermaid theme={null}
graph LR
    A[User Browser] -->|Passive Read| B(Content Script)
    B -->|Sanitized DOM| C{Background Worker}
    C -->|Raw Text| D[LLM Analysis Pipeline]
    D -->|Structured Opportunity| E[Contra Database]
    D -->|Discard| F[Irrelevant]
    
    style A fill:#f9f,stroke:#333,stroke-width:2px
    style E fill:#bbf,stroke:#333,stroke-width:2px
```

### Engineering Spotlight: Invisible Authentication

One of the critical UX requirements was "zero-config". We didn't want users to have to paste API keys or cookies. We implemented a seamless session detection mechanism.

The extension monitors network requests to specific domains (`linkedin.com`, `twitter.com`) to passively detect active sessions.

```typescript theme={null}
// Simplified logic for session detection
chrome.webRequest.onCompleted.addListener(
  (details) => {
    if (details.url.includes("linkedin.com/feed") && details.statusCode === 200) {
      // We have a valid session
      updateSessionStatus('linkedin', true);
      
      // Trigger passive scan of the feed content
      injectContentScript(details.tabId);
    }
  },
  { urls: ["*://*.linkedin.com/*"] }
);
```

> \[!NOTE]
> **Privacy First**: All data extraction happens locally or via ephemeral processing. We strictly adhere to a "read-only" policy regarding user credentials—we never store or transmit session cookies to our servers.

## The Impact

The results were immediate and compounding. Because every new user brings their own unique social graph and IP address, the system scales linearly with user growth.

* **5 Million+ Opportunities Found**: We successfully scaled inventory by 10x, solving the marketplace liquidity problem.
* **30,000+ Active Nodes**: Effectively creating a massive, distributed, residential proxy network for free.
* **High Fidelity**: Because data is sourced from real user feeds, it contains high-intent opportunities often invisible to public scrapers.
* **4.6/5 Star Rating**: By bundling this utility with genuine value for the freelancer (automated job matching), we achieved high retention.

> "Doug operated quickly & efficiently, and even proposed ways to improve the feature to exceed our expectations. 10/10!"
>
> — **Allison Nulty**, Head of Product, Contra

## Why This Matters

This architecture demonstrates a shift from **Centralized Extraction** to **Edge Discovery**. By distributing the workload to the edge (the user's browser), we bypassed the scaling limitations of traditional scraping and built a self-reinforcing data moat.

<Card title="View on Chrome Web Store" icon="chrome" href="https://chromewebstore.google.com/detail/indy-ai/hgpklbhmaaajglapkgebgjjlccndmcoo">
  Join 30,000+ freelancers using Indy.ai
</Card>