Slav Fokin

Tiny World: big data processing and AI-powered chat

Tiny World is a synthetic population platform that generates 10,000 statistically realistic U.S. residents, each modeled as a rich data profile spanning demographics, health, income, politics, religion, and lifestyle. Rather than sampling attributes independently, the platform draws on real U.S. Census microdata (ACS PUMS) to preserve the natural correlations between traits — so income aligns with education, age with occupation, and housing with household size. This grounding in real-world joint distributions makes the synthetic population far more representative than typical dummy data, and useful for genuine social research and simulation. Each persona is further enriched by AI-generated details — taglines, hobbies, backstories — bridging the gap between statistical abstraction and human-feeling data. The web application, built with Next.js and React, lets users explore the population interactively: filtering by demographics, visualizing distributions, and drilling into individual profiles in real time. An AI-powered conversational interface allows researchers to ask natural-language questions about the population and receive context-aware answers grounded in the underlying data. The project integrates multiple real-world datasets beyond the Census, including SSA baby names, surname frequency tables, housing surveys, and national health interview data, to make each synthetic individual as believable as possible. Authentication, subscription management, and a credit system for AI interactions are all production-ready, making Tiny World a full end-to-end product rather than a prototype. The platform sits at the intersection of open government data, statistical modeling, and generative AI — turning raw public datasets into something genuinely interactive and explorable. Tiny World demonstrates how AI and data infrastructure can work together to create tools that make complex population-level information accessible and actionable.