Open ML learning project

Making STEM easier to understand.

Meet AIMS STEM AI Tutor, an experimental AI teaching assistant fine-tuned from Qwen2.5-0.5B-Instruct to explain programming, robotics, electronics, and STEM concepts in friendly language.

Explore the model Learn about the project ↓

Project showcase • Interactive inference is not enabled on this static page.

A hands-on AI learning project

This project explores how a small open-weight language model can be adapted for educational explanations without retraining a large model from scratch.

Beginner-friendly explanations

Focused on understandable, step-by-step responses and simple real-life STEM examples.

Fine-tuned language model

Built on Qwen2.5-0.5B-Instruct and customized using parameter-efficient LoRA training.

Early experiment

24 training examples and 6 held-out examples were used in the initial tutorial experiment. More evaluation is needed.

Tools and technologies

An accessible stack for exploring open models and practical AI development.

PythonQwen2.5-0.5B-InstructUnslothLoRA / QLoRAPyTorchHugging FaceGoogle ColabTesla T4Gradio (experimental UI)

How the model was developed

A small supervised fine-tuning workflow to understand the fundamentals of language model adaptation.

01LoadStart with Qwen2.5-0.5B-Instruct in 4-bit mode.
02PrepareFormat STEM question-and-answer examples as model conversations.
03Fine-tuneTrain LoRA adapters on a Google Colab Tesla T4 GPU.
04Save & shareExport adapter weights to the Hugging Face Model Hub.
Project status: Research and learning prototype

The fine-tuned adapter is publicly available, but the small training set does not establish strong factual accuracy or generalization. This website introduces the project; it does not run the language model. A working hosted chatbot may be added later.

Explore the published model

View the model card, adapter files, and configuration on Hugging Face.

Open model repository ↗