You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
RI RAG is a concept I invented to provide an Infromation Retrieval system that offers:
Less memory usage than standard RAG;
Tracability on retrived content;
Easier and centralized content evolution and update;
The prototype hyrbridising a Mysql database with a Chroma vector have already proven 23,2% of memory footprint reduction but introduce latency. Now adding another system to the loop requires additional memory usage. This phase of the research will be to work on making sure the gain on memory footprint induce on the overall RAM consumption.
On the other hand, RI RAG has been tested on non realistic data and need to be scale on a heavier dataset. This phase of the research will also involve scalability testing.
The project goals are:
coming with a refined architecture that prove the usability of the concept
evaluation how the system scale along with the increase of database records
To achieve them I must :
Generate 2 synthetic dataset of 1k and 10k unique and distinct entries and verify the usability of these datasets
Optimize the database connection and transactions to make sure the used operations are o(1) or near o(1)
The funding will be principally used on:
AI assisted the datasets generation
AI assisted coding and database optimization
Tokens for LLM usage
Human assistanship if applicable
I am alone in my team. This is my first independent resarch since graduation in december 2025 and this RI RAG MVP is already available on Github. I already worked on hybridization for my capstone project where I used together an ES (spam assassin) and AI models (SVM/Random forest) for spam classification using the spam assassin corpus.
Too more confidence in AI. I already generated datasets with AI for the MVP just to discover that it could make mistakes that could drag the entire research. The experience I got with AI assistantship in system design and code generation/optimization revealed that I should be careful and be the entire master and sole brain of the project.
0$.