AI Radar

Verified AI Provider Intelligence & Sentinel Quarantine Engine

An automated market intelligence engine monitoring OpenAI, Anthropic, Gemini, and xAI pricing, models, and lifecycle changes with a Sentinel validation pipeline that prevents broken extractions from corrupting history.

System Visuals & Verification Evidence

1. The Engineering Problem

AI developers and engineering teams make critical architectural decisions based on stale documentation, lagged aggregators, or unstructured scrapers. When an AI provider changes their HTML layout, naive scraping pipelines return zero models and overwrite historical records with empty snapshots. The engineering challenge was building an auditable pipeline that treats scraper outputs as raw observations rather than automatic truth.

2. What Was Built

Built for the WeMakeDevs "Into the Scrape-Verse" hackathon, AI Radar monitors official provider portals across pricing, model catalogs, and lifecycle deprecations. Raw extraction payloads pass through the "Sentinel" engine—a multi-stage validation layer that checks schema contracts. If an extraction fails or is incomplete, the system quarantines the run and retains the last-known-good state.

3. System Architecture & Topology

ASCII Topology

Observation-to-Truth architecture: Bright Data Scraper Studio fetches provider pages; Sentinel applies deterministic contract tests; valid snapshots are committed to canonical history and rendered on a Next.js UI.

+-----------------------------------------------------------------------+
|                      OFFICIAL PROVIDER WEBSITES                       |
|         OpenAI  ·  Anthropic  ·  Google Gemini  ·  xAI               |
+------------------------------------+----------------------------------+
                                     |
                                     v
+-----------------------------------------------------------------------+
|                    BRIGHT DATA SCRAPER STUDIO                         |
|  Dynamic headless web scraping & DOM extraction                       |
+------------------------------------+----------------------------------+
                                     | Raw Observations (JSON)
                                     v
+-----------------------------------------------------------------------+
|                           SENTINEL ENGINE                             |
|  - Source Contract Validation (Required field completeness)           |
|  - Anomaly & Zero-Model Drop Detection                                |
|  - Decision Gate: PASS -> Commit to Canonical History                 |
|                   FAIL -> Quarantine payload & Preserve Last-Good     |
+------------------------------------+----------------------------------+
                                     | Verified Snapshots
                                     v
+-----------------------------------------------------------------------+
|                         AI RADAR WEB (Next.js)                        |
|  - Real-time Pricing Comparison & Token Calculator                    |
|  - Model Capability Matrix & Context Window Tracker                   |
|  - Deprecation & Lifecycle Alerts                                     |
+-----------------------------------------------------------------------+

4. Mohamed's Personal Implementation

Co-architected the Sentinel validation pipeline that distinguishes broken HTML extraction from actual ecosystem model removals.

Constructed the Next.js interactive web interface displaying model capabilities, price-per-token comparisons, and deprecation timelines.

Created structured data contracts defining mandatory fields for pricing tiers, context lengths, and model modalities.

Deployed the application to Vercel and configured public GitHub repository integration.

5. Engineering Challenges & Solutions

Challenge: Preventing False-Positive Ecosystem Deprecations

Solution: Enforced a Sentinel quarantine rule: if a scraper returns fewer models than a predefined tolerance threshold compared to the previous snapshot, the payload is held for review and the frontend continues serving the verified last-known-good dataset.

Impact: Eliminated corrupted historical charts and prevented false alarms for developers.

6. Technology Stack

Frontend & API

Next.js App RouterTypeScriptTailwind CSSVercel

Data Harvesting

Bright Data Scraper StudioWeb Scraping APIs

Data Integrity

Sentinel Validation EngineJSON Schema Contracts

7. Verified Outcomes

•Awarded Top 50 in the global WeMakeDevs "Into the Scrape-Verse" Hackathon.
•Fully deployed and verified at https://ai-radar-orpin.vercel.app with public repository and video demo.

Visit Live Platform Source Code Watch Demo

← Back to Portfolio