Projects
Isgen.ai
AI Detection Engine | Content Classification | LLM Evaluation | API SaaS
Undetectable AI Essay Writer
ISGEN.ai is an AI detection platform designed to evaluate whether content is AI-generated, paraphrased, or human-written using classification and probability-based scoring. The platform includes text and image authenticity detection, enabling users, institutions, and platforms to validate originality. It is engineered as a public API solution, enabling developers to integrate AI detection into their SaaS products or writing platforms.
Challenge
Role
AI & Backend Engineer responsible for designing the detection logic, implementing classification workflows, creating probability scoring systems, and building API endpoints accessible via RapidAPI.
Project Goals
- Detect AI Text Reliably – Provide high-confidence classification of human vs AI-generated content.
- Identify Paraphrased AI Content – Mitigate bypasses by paraphrasing and rewriting.
- Offer Simple Developer Integration – Expose detection capabilities through public APIs.
- Support Image Authenticity Detection – Check whether visual assets were generated or manipulated by AI.
Development Process
Data Collection | Data Pre-processing | LLM Fine-tuning
Scraped diverse human-written content using Selenium and BeautifulSoup. Cleaned data for consistency and balance across writing styles.
Fine-tuned GPT-3.5-turbo to mimic human writing patterns. Tested against AI detection tools, adjusting for human-like variability without sacrificing coherence.
Stack Used:
Final Outcome
The system features a Text Detection Pipeline that analyzes linguistic patterns and semantic irregularities, supported by a Probability‑Based Scoring System that provides confidence levels instead of binary outputs. Its Image AI Detection Logic identifies AI‑generated or manipulated visual artifacts with high precision. A publicly deployed API on RapidAPI enables global developer access, while robust backend and cloud integration ensure stable, scalable performance suitable for production‑level consumption.
