⚙️Technical Architecture & Integration

Overview

This paper describes level architecture of the ICanProveIt application. ICanProveIt architecture is designed to support the following main use cases:

1. Issuance of Proof-of-Learning Certificates

2. Verification of Proof-of-Learning Certificate

3. Uploading of documents into our document repository

4. Composition of exam for tests

5. Proctoring exam

6. Create report

Implementation of the architecture

More information is available in the technical section of our Whitepaper.

pageDominium-Stellar-Integration

📃pageWhite Paper

Component	Name	Description
1	Dominium	Dominium is an open source highly scalable product of FuixLabs and a full implementation of the DID standard for decentralized ID.
2	DID Based education certificates.	Certificates that conform to the VC-EDU standard. Our proposers have integrated with 15 universities as part JFF Plugfest
3	LLM	A Large Language Model such as ChatGTP used to generate questions.
4	ReactJS front end	We will support a professionally designed front end implemented using ReactJS
5	LangChain	LangChain is used from tuning and Prompt Engineering
6	NodeJS	NodeJS is a well known server technology. FuixLabs has build applications that scale up to millions of users using NodeJS technology.
7	Document Repository/Vector Database	Vector database to store custom embeddings.
8	DID based id for learners	A DID based on the W3.org standard.

diagram 1

Creating a Question/Answer

Create a Corpus of Relevant Documents

In order to generate multiple unique exams and ensure the variety of question/answers, we will maintain a corpus of commons licensed document that are first verified by qualified academics and domain specialists.

1. Someone uploads a document to a temporary storage location.

2. The document is approved by reviewers for licensing and content.

3. Embeddings are extracted and added to our vector database.

Contnet Upload

Creating An Examination for The Learner

1. Learner uploads materials that they have covered.

2. The uploaded documents are examined for testable content.

3. Questions are generated from the content in the document repository and an examination is created.

Exam is Virtually Proctored

1. The learner must have a video camera on their machine. The exam uses special AI technology similar to technology used for KYC to make sure the user does not try to fake image.

2. After the user answers the questions, which can be multiple choice, written essay, or code, the exam is graded using GenAI.

3. A certificate is generated and hosted on our site.

4. The learner can now add his certificate to his resume, post it on social media...

LLM Architecture

Exam Creation and grading is accomplished with multiple LLM agents each tuned to a specific responsibilities. Agents collaborate to deliver the ICanProveIt exam generation, grading, and topic extraction. The Stellar prototype will have 2 agents with more added later.

LLM Tuning

LLM Agents will implement a modular RAG architecture that allows continuously updating as new documents are added.