AI Congress Online — Multi-Model Cross-Validation Platform for Reproducible Data Cleaning
A live public AI platform that runs multiple LLMs on the same data-cleaning task and cross-validates their outputs, making AI-assisted pipelines reproducible and audit-friendly.
Business Challenge
Single-model AI pipelines are hard to verify: outputs drift between runs and between models, and data-cleaning steps cannot be re-audited after the fact. Organisations evaluating AI tooling need reproducible, cross-validated results instead of opaque single-model answers.
Solution Delivered
Built AI Congress Online, a live multi-model platform that analyses the same input with several LLMs and returns a cross-validated consensus answer with explicit disagreement flags for human review. Each run is reproducible and traceable, so data-cleaning and analysis steps can be re-audited. The product is publicly accessible for demonstration (aicongressonline.com) and its source is published on GitHub (github.com/kinsin818/AI-Congress-Online), letting organisations verify the implementation themselves.
Outcomes Achieved
Live and publicly verifiable platform; reproducible multi-model runs with cross-validation; open-source implementation for independent review. Used as the flagship reference for an AI full-stack consulting profile.
Measurable Business Outcome
Live and publicly verifiable platform; reproducible multi-model runs with cross-validation; open-source implementation for independent review. Used as the flagship reference for an AI full-stack consulting profile.