Problem
Toxic content detection usually relies on resource-heavy machine learning models that can be slow, resource-intensive, and hard to run on lightweight edge servers.
AI / Automata Content Moderator Project Build
Detox
Web-based automata engine for toxic content detection using NFA slang normalization and deterministic DFA matching.
Role
Backend & DFA Engine Developer
Year
2026
Stack
FastAPI + Python
Solution
Detox uses formal language theory (NFA and DFA) to normalize spelling and match toxic patterns deterministically in O(n) time, making it incredibly fast and lightweight.
Overview
What I Built
Detox is an automata-based content moderation system built for toxic content detection. It utilizes a three-layer processing engine: NFA for normalizing slang/spelling variations, regular expressions for defining formal patterns, and DFA execution for deterministic O(n) toxic word matching.
Technical Highlights
Engineering Notes
Deterministic DFA matching executing in O(n) time complexity
FastAPI backend structure with custom lexicon parsing
Slang normalization mapping NFA state steps
Interactive bulk text checking dashboard