B.L.A.S.T.: Business-Idea Intelligence Engine

01 Problem and context

Selecting a business idea often relies on hunches without empirical evidence. There is a lack of structured methods to adapt documented success stories to the specific pain points and context of the LATAM market.

Business intelligence applied to the analysis of documented case studies for opportunity development in the LATAM market.

02 Solution

B.L.A.S.T. automates data extraction from YouTube and executes a five-stage inference pipeline using Google Gemini. It maps real case data against LATAM pain points and user RPM profiles to output validated proposals traceable to source videos.

1. 01Extraction & Ingestion: scraping YouTube transcripts and metadata via yt-dlp and YouTube Data API v3 into PostgreSQL.
2. 02Modeling & Integration: defining schemas to incorporate LATAM pain points and user RPM profiles.
3. 03Inference Pipeline: implementing the five reasoning stages (Blueprint, Link, Architect, Stylize, Trigger) powered by Google Gemini.
4. 04Generation & Traceability: producing business proposals linked directly to documented source cases.
5. 05Automation: configuring scheduled update tasks using APScheduler.

03 How it's built

An intelligence system that scrapes YouTube to extract transcripts and metadata from Starter Story success cases. It processes data through a five-stage pipeline (Blueprint, Link, Architect, Stylize, and Trigger) powered by Google Gemini. The engine cross-references extracted cases with LATAM market pain points and the user's RPM profile. It generates business proposals traceable to their original source cases. Scheduled ingestion via APScheduler keeps the underlying PostgreSQL database on Supabase current.

04 Ethical decisions

• ✦Data traceability: every proposal includes explicit references to the source cases used.
• ✦Intellectual property protection: source code maintained in a private repository.

05 Highlights

• ✦Structured five-stage reasoning pipeline: Blueprint, Link, Architect, Stylize, and Trigger using Google Gemini.
• ✦Automated transcript and metadata scraping via yt-dlp, YouTube Data API v3, and youtube-transcript-api.
• ✦Full traceability linking generated proposals directly back to original Starter Story source cases.
• ✦Scheduled data ingestion via APScheduler to keep the PostgreSQL database on Supabase updated.
• ✦Multivariable data mapping incorporating LATAM market context and user RPM profiles.

06 Limitations

The dataset originates from public YouTube scraping of Starter Story videos (transcripts and metadata) and does not constitute audited financial data.