pinned
Running on Zero
Agents
Nalanda-7B: JEE & NEET Question Solver
🎓
Solve JEE/NEET questions with a GRPO-trained 7B model
AI & ML interests
Indian STEM education AI: benchmarks, multimodal models, document understanding, and multilingual (EN/HI) datasets for tutoring and assessment.
Recent Activity
View all activity
Organization Card
Nalanda Data
Open data and models for Indian STEM education AI.
We build and publish datasets and fine-tuned models grounded in Indian academic content — JEE & NEET competitive exam questions, school and college textbooks, and multimodal science material. The goal is to make Indian education AI workable for researchers, builders, and edtech companies who can't easily source this data elsewhere.
🌐 nalandadata.ai
📧 info@nalandadata.ai (commercial) · tech@nalandadata.ai (technical)
What we publish
Datasets
| Dataset | Focus | License |
|---|---|---|
NalandaJEENEETBench |
JEE & NEET benchmark across Physics, Chemistry, Mathematics, Biology | CC-BY-NC-4.0 |
DrishtiTable |
Table structure recognition in Indian textbooks (EN + HI) | Apache-2.0 |
nalanda-image-qa |
1,000 multimodal STEM Q&A pairs (image + text) | CC-BY-4.0 |
Models
| Model | Built on | Use case | License |
|---|---|---|---|
nalanda-qwen-7b-grpo |
Qwen 2.5 7B Instruct + GRPO | JEE / NEET problem solving | Apache-2.0 |
nalanda-image-vl |
Llama 3.2 11B Vision (LoRA) | Multimodal STEM Q&A | Llama 3.2 |
DrishtiTable-Qwen2.5-VL-7B |
Qwen 2.5 VL 7B (LoRA) | Table structure recognition | Apache-2.0 |
Demos
nalanda-jee-neet-solver— try the JEE/NEET solver in your browser.
Who this is for
- Researchers building or evaluating models on Indian-context STEM
- Edtech companies training tutoring, grading, or content-generation systems
- AI labs that need benchmarks reflecting non-Western curricula and multilingual (EN / HI) educational content
Data and licensing
The public artifacts on Hugging Face are samples of larger internal datasets. Each repo has its own license, listed in its dataset/model card.
- Open-licensed releases (Apache-2.0, CC-BY-4.0) can be used commercially under the terms of those licenses.
- Non-commercial releases (CC-BY-NC-4.0) require a separate commercial license for production or revenue-generating use.
- Full-scale versions, custom slices, and licensed access to the parent corpora are available on request.
For commercial licensing, full dataset access, custom data work, or partnerships:
For technical questions, integration help, or fine-tuning support: