Autonomous AI system that fetches research papers from PubMed, performs RAG-based analysis, generates novel hypotheses, designs experiments and produces publication-style research proposals — with zero human intervention.
Topic Input
→ 🔍 PubMed Fetch (35M+ papers)
→ 🧠 RAG Analysis (FAISS + SentenceTransformers)
→ 💡 Hypothesis Generation (Groq LLaMA 3.1)
→ 🧪 Experiment Design
→ 📝 Research Proposal Output
| Step | Feature | Description |
|---|---|---|
| 1 | PubMed Search | Auto-fetches top papers for any topic |
| 2 | Literature Analysis | RAG pipeline summarizes key findings |
| 3 | Hypothesis Generation | LLM generates novel, testable hypotheses |
| 4 | Experiment Design | Dataset, methodology, timeline, controls |
| 5 | Research Proposal | Publication-style output with peer review |
| Component | Technology |
|---|---|
| Paper Source | PubMed API (35M+ papers, no rate limit) |
| Vector Store | FAISS |
| Embeddings | SentenceTransformers (all-MiniLM-L6-v2) |
| LLM | Groq LLaMA 3.1 (8B Instant) |
| UI | Gradio |
# Install dependencies
pip install faiss-cpu sentence-transformers groq gradio requests
# Set your Groq API key (free at console.groq.com)
export GROQ_API_KEY="your_key_here"
# Run
python app.pyOr open directly in Google Colab using the badge above.
groq
faiss-cpu
sentence-transformers
gradio
requests
Input topic: large language models medical imaging
Output includes:
- 5 fetched papers from PubMed (2024–2025)
- RAG-based literature summary
- 3 novel hypotheses
- Full experiment design with methodology
- Publication-style research proposal
- 🤗 Live Demo: huggingface.co/spaces/Vi-bha/ResearchMind
- 🔍 PaperLens: huggingface.co/spaces/Vi-bha/PaperLens
- 🔬 MedLens: huggingface.co/spaces/Vi-bha/MedLens
Built by Vibhavari Tummewar | MTech Advanced Computing, MANIT Bhopal