GeoGov Lite

Project Overview

The project is a prototype system addressing the TikTok TechJam 2025 problem statement focused on geo-specific compliance logic detection. Using advanced large language models (LLMs) and AI capabilities, the system flags features requiring regional legal guardrails to proactively mitigate regulatory risks. It generates auditable, automated evidence proving features were screened for compliance needs to streamline governance and audit readiness.

Features and Functionality

Automatically analyzes feature artifacts (titles, descriptions, documents) to detect geo-specific compliance requirements, such as country-specific copyright or child protection laws.
Flags features that require geo-specific compliance logic with clear reasoning.
Optionally associates flagged features with relevant regulations.
Provides traceable, auditable outputs to enable swift regulatory inquiry responses.
Designed to reduce manual compliance governance efforts and legal exposure.
Supports human evaluation for ambiguous cases to improve accuracy.
Implements mechanisms to fill LLM context gaps by integrating domain-specific knowledge and specialized detection techniques.

Development Tools Used

VSCode: Used as the primary code editor for development.
Docker: Utilized for containerizing the application to ensure consistent environments across development and deployment.

APIs Used

OpenAI API: Provides access to advanced natural language processing models for feature artifact analysis and compliance detection.
Congress API: Used for accessing legislative data and member information to enrich compliance logic where applicable.

Assets Used

TikTok Dataset: A synthetic dataset representing feature artifacts for testing and validation.
Additional libraries: React for frontend UI development, FastAPI for building backend API services, and CrewAI for multimodal AI capabilities.

Libraries Used

React: JavaScript library for building modern user interfaces.
FastAPI: High-performance Python framework for web APIs.
OpenAI SDK: For integrating AI model functionalities.
CrewAI SDK: Multi-modal AI platform tools.

Problem Statement

Build a prototype system that utilizes LLM capabilities to flag features requiring geo-specific compliance logic, turning regulatory detection from a blind spot into a traceable, auditable output. The solution empowers TikTok to implement legal guardrails before feature launches and confidently respond to regulatory inquiries with automated traceability.

Dataset

The project uses a synthetic TikTok dataset provided for the challenge. No real business data is involved.

Repository Information

The public GitHub repository containing the project source code and documentation, including setup instructions, is available at: https://github.com/bchewy/taktim

The .env require certain API keys, namely:

PINECONE_API_KEY
OPENAI_API_KEY
EXA_API_KEY
PERPLEXITY_API_KEY

US-Specific

CONGRESS_API_KEY
GOVINFO_API_KEY

Demonstration Video

A demonstration video shows the project functioning on the intended platform. The video is under 3 minutes, includes live usage of the system, and is publicly visible on YouTube. https://youtu.be/CDNeyueEYYg