GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
r-karra
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
r-karra
Model Tree
r-karra/Gemma-2-9B-JEE-Socratic-Final
Input Modalities
Output Modalities
Supported Functionality
This project focuses on building an AI-driven educational assistant specifically for students preparing for the IIT-JEE Advanced examination.
Unlike conventional AI tutors that directly provide answers, this model follows the Socratic Method, encouraging students to think critically by asking carefully designed questions that guide them toward the solution.
The goal is to promote:
The tutor acts as a learning facilitator, not merely an answer generator.
For every student question, the model:
text
.├── Dataset│ ├── JEE Advanced Papers│ ├── Chemistry Syllabus│ ├── Physics Notes│ └── Mathematics Problems│├── Data Processing│ ├── Cleaning│ ├── Formatting│ └── Socratic Prompt Creation│├── Fine-Tuning│ ├── Unsloth│ ├── Hugging Face Transformers│ └── Gemma-2-9B│├── Model│ └── Gemma-2-9B-JEE-Socratic-Final│└── Inference └── Hugging Face TransformersThe dataset consists of carefully curated IIT-JEE Advanced educational resources.
Subjects include:
Additional curated educational resources include:
These resources help improve conceptual understanding and expand the tutoring capability beyond previous examination questions.
The project uses Unsloth for efficient GPU-based fine-tuning.
Supported environments:
Benefits:
The original JEE Advanced questions are transformed into conversational teaching examples.
Each example teaches the model to:
The Gemma model is instruction-tuned to recognize pedagogical patterns.
Instead of:
"Here is the answer."
The model responds like:
"Which thermodynamic quantity determines spontaneity?"
"What happens to ΔG when entropy increases?"
This encourages active learning.
After fine-tuning, the model is published on Hugging Face.
Model Repository
text
r-karra/Gemma-2-9B-JEE-Socratic-FinalYou do not need to convert the model to GGUF format.
The model can be loaded directly using the Hugging Face transformers library.
bash
pip install transformers accelerate torchpython
from transformers import AutoModelForCausalLM, AutoTokenizerimport torchfrom google.colab import userdata # Configurationmodel_id = "r-karra/Gemma-2-9B-JEE-Socratic-Final"token = userdata.get("HF_TOKEN") device = "cuda" if torch.cuda.is_available() else "cpu" # Load tokenizerprint("Loading tokenizer...")tokenizer = AutoTokenizer.from_pretrained( "google/gemma-2-9b-it", token=token) # Load modelprint("Loading model...")model = AutoModelForCausalLM.from_pretrained( model_id, token=token, device_map="auto", torch_dtype=torch.float16).to(device) def ask_tutor(question): prompt = ( "System: You are an expert IIT-JEE Advanced tutor " "using the Socratic method.\n" f"User: {question}\n" "Model:" ) inputs = tokenizer(prompt, return_tensors="pt").to(device) with torch.no_grad(): outputs = model.generate( **inputs, max_new_tokens=500 ) response = tokenizer.decode( outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True ) return response print( ask_tutor( "How can I determine if a reaction is spontaneous using Gibbs free energy?" ))How can I determine whether a reaction is spontaneous using Gibbs free energy?
Before thinking about spontaneity, can you recall the mathematical expression relating Gibbs free energy to enthalpy and entropy?
What does a negative value of ΔG imply about the direction of a process?
How would increasing temperature affect the entropy contribution?
Rather than replacing teachers, this project aims to become an intelligent learning companion that:
The emphasis is on learning the process, not memorizing the answer.
Model Name
text
Gemma-2-9B-JEE-Socratic-FinalHugging Face Repository
text
r-karra/Gemma-2-9B-JEE-Socratic-FinalAll IIT-JEE Advanced question papers remain the intellectual property of their respective examination authorities.
This repository is intended solely for:
No copyright over the original examination content is claimed.
Special thanks to:
for providing the tools and ecosystem that made this project possible.
If you find this project useful, please consider starring the repository and citing it in your educational or research work.
text
@misc{jee-socratic-tutor, author = {Rajesh Kumar Karra}, title = {Gemma-2-9B JEE Advanced Socratic Tutor}, year = {2026}, publisher = {Hugging Face}, model = {r-karra/Gemma-2-9B-JEE-Socratic-Final}}.├── Dataset│ ├── JEE Advanced Papers│ ├── Chemistry Syllabus│ ├── Physics Notes│ └── Mathematics Problems│├── Data Processing│ ├── Cleaning│ ├── Formatting│ └── Socratic Prompt Creation│├── Fine-Tuning│ ├── Unsloth│ ├── Hugging Face Transformers│ └── Gemma-2-9B│├── Model│ └── Gemma-2-9B-JEE-Socratic-Final│└── Inference └── Hugging Face Transformersfrom transformers import AutoModelForCausalLM, AutoTokenizerimport torchfrom google.colab import userdata # Configurationmodel_id = "r-karra/Gemma-2-9B-JEE-Socratic-Final"token = userdata.get("HF_TOKEN") device = "cuda" if torch.cuda.is_available() else "cpu" # Load tokenizerprint("Loading tokenizer...")tokenizer = AutoTokenizer.from_pretrained( "google/gemma-2-9b-it", token=token) # Load modelprint("Loading model...")model = AutoModelForCausalLM.from_pretrained( model_id, token=token, device_map="auto", torch_dtype=torch.float16).to(device) def ask_tutor(question): prompt = ( "System: You are an expert IIT-JEE Advanced tutor " "using the Socratic method.\n" f"User: {question}\n" "Model:" ) inputs = tokenizer(prompt, return_tensors="pt").to(device) with torch.no_grad(): outputs = model.generate( **inputs, max_new_tokens=500 ) response = tokenizer.decode( outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True ) return response print( ask_tutor( "How can I determine if a reaction is spontaneous using Gibbs free energy?" ))@misc{jee-socratic-tutor, author = {Rajesh Kumar Karra}, title = {Gemma-2-9B JEE Advanced Socratic Tutor}, year = {2026}, publisher = {Hugging Face}, model = {r-karra/Gemma-2-9B-JEE-Socratic-Final}}