Global Image Dataset, Web Scraping

A remote engagement with Limitless Capital (Singapore) building the upstream pipeline that feeds computer-vision models, emphasising geographic diversity, licensing hygiene, and labelling consistency.

The brief.

AI inclusivity starts at the dataset. Limitless Capital needed image corpora that genuinely represented users across continents, not the usual North-American skew. The role combined engineering, curation, and quality assurance under remote, performance-tracked conditions.

How it was built.

1. STEP 01Wrote modular Python scrapers covering 50+ sources with respectful rate limiting and retry logic.
2. STEP 02Curated images across 30+ countries, balancing demographic and contextual diversity.
3. STEP 03Authored data-quality standards adopted by the cross-functional team.
4. STEP 04Operated independently against weekly performance benchmarks in a remote environment.

What it moved.

• →Collected 15,000+ images while cutting manual collection time by 65%.
• →Improved AI inclusivity metrics on downstream models by 28%.
• →Established the QA checklist that became the team's onboarding document.

Tools used.