EduCore AI is an education management platform I built to show how three kinds of AI can work together in one product: prediction, document search and agents. It is my own project, not a client engagement.
The problem
Schools and training providers face the same three time sinks. Staff find out too late that a student is struggling. The same policy questions come in again and again. And admissions involves reading, checking and sorting applications by hand.
What I built
At-risk student prediction. A Random Forest model scores each student’s risk of falling behind. On its own, a risk score isn’t useful to a teacher, so each prediction comes with a SHAP explanation showing which factors pushed the score up or down. Staff can see why a student was flagged, not just that they were.
Policy chatbot (RAG). Students and staff ask questions in plain English. The chatbot searches the institution’s own policy documents with FAISS, then answers with LLaMA 3 running on Groq, using only what it found. That keeps answers tied to the real policies instead of a model’s general knowledge.
Admissions pipeline with three agents. A CrewAI pipeline splits admissions work between three agents, each with a narrow job, so no single agent has to do everything.
Design choices
- Explainable before accurate. In education, a prediction nobody can explain won’t be trusted or acted on. SHAP was part of the design from the start.
- Answers from documents only. The chatbot is grounded in retrieved text, which is the most practical way to keep an LLM from making up policy.
- Small agents with narrow roles. Three focused agents are easier to test and debug than one agent that does it all.
Where this applies to your business
The same building blocks work outside education: prediction to spot customers or accounts at risk, a document chatbot for your policies and FAQs, and agents to process applications, orders or intake forms.
Frequently asked questions
What does EduCore AI do?
It combines three kinds of AI in one education platform: a model that flags students at risk of falling behind, a chatbot that answers questions from the institution’s own policy documents, and a three-agent pipeline that handles admissions work.
How does EduCore AI explain its at-risk predictions?
A Random Forest model scores each student, and every score comes with a SHAP explanation showing which factors pushed it up or down, so staff can see why a student was flagged.
Which technologies is EduCore AI built with?
A Random Forest model with SHAP explanations for predictions, FAISS for document search, LLaMA 3 on Groq for the policy chatbot, and CrewAI for the admissions agents.
Can the same approach work outside education?
Yes. The same building blocks can predict customers or accounts at risk, answer questions from your policies and FAQs, and process applications, orders or intake forms.
Try it free: take the 3-minute AI readiness quiz, see what missed calls cost you with the AI savings calculators, try the in-browser RAG chatbot demo or listen to the AI receptionist demo.
Related services: RAG chatbot development · Multi-agent systems · Machine learning consulting
Book a free consultation to talk through your use case.
AI/ML engineer building LLM agents, RAG systems and automation for growing businesses.
