Case studies

Unified portal containing comprehensive engineering evaluations, architectural audits, and trade-offs.

Select case study

⚡

StartupWire

The Autonomous AI-Powered Tech Startup News Platform

1. Overview

StartupWire was designed to solve the clutter and human latencies in modern startup news curation. By utilizing Google's Gemini API to filter noise and programmatically index relevant articles, StartupWire achieves automated, high-quality publication with zero human intervention.

2. Business Problem & Targets

Founders, developers, and investors face information overload. Major news portals are often slow to report niche updates, manually summarizing tech breakthroughs is extremely time-consuming, and programmatic content engines typically lack human-like verification or produce repetitive, duplicate topics.

3. System Architecture & Relational Specs

Supabase PostgreSQL. Structured schemas for articles, feed sources, and tags. Integrated a pgvector module to generate text embeddings for article titles, enabling strict cosine similarity lookups to prevent duplicate publication.


+------------------+     +------------------------+     +-------------------+
|  RSS Feed Nodes  | --> | Node.js Scraper Worker | --> | Gemini 1.5 Flash  |
| (Tech Crunch, HN)|     | (Extract Content/URL)  |     | (Topic Curation)  |
+------------------+     +------------------------+     +-------------------+
                                                                  |
                                                                  v
+------------------+     +------------------------+     +-------------------+
|   Client App     | <-- | Supabase PG (RLS Node) | <-- | Vector Distance   |
| (Edge Cached V4) |     |  (Structured JSON)     |     | (Deduplication)   |
+------------------+     +------------------------+     +-------------------+

4. Security Protocols

Configured strict Content Security Policies (CSP), HTTP-only SameSite cookies, input data schemas sanitization via Zod validators, and granular PostgreSQL Row Level Security (RLS) policies.

5. SEO & Performance Tunings

Programmatic SEO framework. Dynamic sitemap generation, structured JSON-LD NewsArticle schemas, localized meta-descriptions (<160 chars), custom og:image layouts, and clean routing slugs.

6. Challenges & Lessons

LLM Hallucinations and duplicate feeds. Solved by writing strict system instructions for the Gemini model, validating JSON output formatting, and maintaining vector indexes to filter duplicate topics.