GENESIS.AI

#cloudflarechallenge #devchallenge #ai #generativeai

This is a submission for the Cloudflare AI Challenge.

What I Built

Our app, Genesis.AI (stands for 'Generative Analysis AI') harnesses capabilities of Generative AI🤖 to analyze a website's contents along with a summary and analysis report📋 and briefs about the analytical statistics🧮 about the content being safe for digital consumption or not.

Demo

Deploy Link - https://genesis-ai-lake.vercel.app

Sample HTML Website Content Analysis UI-

Sample Target HTML Website UI-

My Code

GitHub Repo. Link- https://github.com/Alphax16/genesis.ai

Journey

We started off keeping in mind the business use-case of Cloudflare itself, some of which are CDN Services, Cloud and Cyber Security, Risk Mitigations and Domain Registration Services.

Our aim was to come up with something which could prove it's importance before the real world problems tackled by Cloudflare on a day-to-day basis.

And, so, we emerged with our idea of developing a Generative AI powered application, Genesis.AI (Generative Analysis AI)

As we very know that Cloudflare and many such similar big tech service based companies offer their services, but soon after releasing their solutions they get detached from their clients and lose their control over how their softwares solutions are being utilized, which many a times leads to serious legal consequences.

Eureka, glowed our light bulbs and we came up with our idea of developing a pilot project of developing and AI based content analysis system, capable of scanning a complete website and it's associated domains and performing a multi-media (including both text and image) analysis and flagging whether the application making use of ones' services is a legitimate and ethical one or not.

For achieving our object, we primarily made use of the following 3 popular AI model workers offered by Cloudflare-

LLAMA-2-7b-chat-fp16
Distilbert-SST-2-int8
ResNet-50

Multiple Models and/or Triple Task Types
The app makes use of Clouflare's Workers AI models for achieving its task (triple task type)-

LLAMA-2-7b-chat-fp16 developed by Meta Platforms, Inc.
Model Description - This model is trained on 2 trillion tokens, and by default supports a context length of 4096. LLAMA 2 Chat models are fine-tuned on over 1 million human annotations, and are made for chat.
Distilbert-SST-2-int8 developed by researchers at Huggingface.
Model Description - The model distilbert-sst-2-int8 refers to a quantized version of the DistilBERT model fine-tuned on the Stanford Sentiment Treebank (SST-2) dataset. The original DistilBERT model was developed by researchers at Hugging Face.
DistilBERT itself is a smaller, faster, cheaper version of BERT (Bidirectional Encoder Representations from Transformers), originally developed by Google. DistilBERT was designed to provide a lighter transformer model that retains most of the performance of BERT but with fewer parameters, making it more efficient to use in terms of both memory and speed.
ResNet-50 developed by Microsoft Research.
Model Description - ResNet50 is a deep Convolutional Neural Network (CNN) architecture that was developed by Microsoft Research in 2015. It is a variant of the popular ResNet architecture, which stands for “Residual Network.” The “50” in the name refers to the number of layers in the network, which is 50 layers deep.

Thus, the project made use of the above mentioned 3 popular AI model AI workers.