First Ascent by
Arisara Weeranarawat
The Quantum Entanglement RAG
LLM Engineering / Retrieval-Augmented Generation
The Proposed Route
An ambitious infrastructure route: building a RAG-powered chatbot for quantum entanglement Q&A using Deepseek R1 distill Llama 70B (4-bit quantized) on an RTX A6000. The climber fetches scientific literature via the Springer Nature API, converts XML to JSON, and implements retrieval-augmented generation with a Dockerized deployment.
🧗 The Crux
This is an LLM engineering project rather than a biochemistry ML route — the connection to the course content (molecular ML, protein science) is thin. Running a 70B model even at 4-bit quantization is hardware-intensive. RAG implementation has many moving parts (chunking, embedding, retrieval, generation) that can each fail independently.
⚠️ Pre-Climb Checklist
✅ Springer Nature API already working — data pipeline in place. ✅ Docker containerization is good engineering practice. ⚠️ 70B at 4-bit still needs ~35GB VRAM — verify A6000 can handle inference. ⚠️ RAG has many moving parts (chunking, embedding, retrieval, generation) — test each component independently.
Guidance
- The RAG architecture itself is valuable — focus on evaluation (retrieval accuracy, answer quality)
- Document the chunking strategy and embedding model choices
- Include example queries and retrieved passages in the final notebook
Source proposal: Arisara_Weeranarawat_FinalProjectProposal.pdf
CHEM 169/269 · Applied AI & Machine Learning for Biochemistry