PaperQA2 is a useful reminder for scientific RAG: leave an evidence trail
A SciencesLoop technical note on PaperQA2, citation precision, and reviewable scientific RAG artifacts.
SciencesLoop · AI research intelligence · AI for Science
SciencesLoop
An AI for Science research intelligence workspace by Lu Zhang.
Tracking AI agents, scientific workflow automation, selected AI labs, and AI for Science startups — then turning the signals into cited notes, product ideas, and small agent prototypes.
Daily AI research radar Scientific agent workflows Cited sources when available Public-safe corpus
Daily AI Research Radar
A source-based working workflow for tracking selected AI labs, AI for Science startups, research preprints, and scientific-agent workflow signals.
Agent interfaceScientific Workflow Agent
A public-safe site agent that retrieves evidence, routes visitors to the right source, and exposes next actions instead of acting like a static portfolio search box.
Build trackAI for Science Lab
Small prototypes around RAG, evaluation, traceable decision support, redox-flow formulation, and scientific workflow automation.
AI & engineering
All projects → lu-zhang-site
lu-zhang-site
This site — academic homepage with structured data layer
Astro static site driven by curated JSON data extracted from a master research record. 99 publications (97 SCI + 2 book chapters), 28 patents (25 issued + 3 applications), 51 conference presentations (38 abstracts + 13 invited talks).
Live
meditation-study
meditation-study
Personal meditation learning platform
Web app for studying a meditation teacher-training course. Built around session JSON, polished script generation, and TTS audio playback. Daily-use product.
Live
guangnian-energy-agent
guangnian-energy-agent
光年新能源智能体 — RAG assistant demo for new-energy content
Bilingual RAG assistant demo/prototype built around Azure OpenAI and Azure AI Search patterns, with a Python aiohttp backend and lightweight web UI. Used as a practical AI x energy demo and workflow prototype.
Live demo
rfb-formulation-agent
rfb-formulation-agent
Redox-flow-battery formulation support prototype
A pre-v0.1 vertical-domain AI agent prototype for redox-flow-battery formulation support. It tests whether structured retrieval, reasoning, and validation workflows can help organize flow-battery formulation knowledge.
In progress · pre-v0.1
AI research radar
All notes →ChemGraph is a better first test for scientific agents
A SciencesLoop note on ChemGraph, computational chemistry agents, and why traceable workflows matter more than fluent chat.
Agentic Scientific Workflows: From Search to Traceable Decisions
A working model for scientific agents that retrieve evidence, plan the next step, preserve uncertainty, and leave a decision trail.
Featured research
All publications → J. Mater. Chem. A — under revision
2026
2026
From machine learning to chemical insight: Darwinian chance and the stability of charge carriers in flow batteries
ML-driven insight into the stability of charge carriers in flow batteries. Submitted to JMCA (TA-ART-01-2026-000771); a direct extension of the JCESR DMB redoxmer line. With Robertson, Shkrob, Assary, J. Zhang.
ACS Applied Energy Materials
2025
2025
Automation-Accelerated Electrolyte Design Mitigates Solubility Competition between Redox-Active Molecules and Supporting Salts
2025 follow-up to the Nature Comm robotic platform — automated HTE + ML for paired BTZ/LiTFSI solubility (~3 M, surpassing the 2 M ceiling). With Robertson, Doan, Han.
Journal of Materials Chemistry A
2025
2025
A comparative computational and SECM study of factors influencing electron transfer at hydrogenated and pristine graphite-propylene carbonate interface
2025 SECM + computational study showing H-functionalization on graphitic electrodes increases electron-transfer kinetics. With Curtiss, Assary, Counihan, Rodríguez-López, J. Zhang.
An integrated high-throughput robotic platform and active learning approach for accelerated discovery of optimal electrolyte formulations
Robotic platform + ML for electrolyte discovery — bridging chemistry and AI.
Angewandte Chemie
2015
2015
Radical Compatibility with Nonaqueous Electrolytes and Its Impact on an All-Organic Redox Flow Battery
Most-cited paper. Foundational radical-cation work for non-aqueous flow batteries.
Energy & Environmental Science
2012
2012
Molecular engineering towards safer lithium-ion batteries: a highly stable and compatible redox shuttle for overcharge protection
First-author EES paper on overcharge-protection redox shuttles — basis for R&D 100 Award (2014).
Scientific Reports
2016
2016
The lightest organic radical cation for charge storage in redox flow batteries
Systematic molecular trimming to find the smallest stable radical-cation core. Featured in Argonne Today.
Recent
latest firstCo-authored manuscript (lead: Lily Robertson; co-authors: Ilya Shkrob, Rajeev Assary, Jingjing Zhang) under revision at Journal of Materials Chemistry A (TA-ART-01-2026-000771). Bridges ML-driven insight to charge-carrier stability in non-aqueous flow batteries — direct extension of the JCESR DMB redoxmer line.
ML-guided HTE for paired BTZ/LiTFSI solubility — m-xylene/acetonitrile mixture dissolves ~3 M of both, surpassing the 2 M ceiling for neat acetonitrile. With Lily Robertson, Hieu Doan, Kee Sung Han.
H-functionalization on graphitic electrodes increases electron-transfer kinetics in non-aqueous redox media. With Curtiss, Assary, Counihan, Rodríguez-López, J. Zhang.
A pre-v0.1 AI prototype testing whether structured retrieval, reasoning, and validation workflows can help organize redox-flow-battery formulation knowledge.
Working on high-level enterprise data/AI workflow tooling at AAOS; public details intentionally limited.
Nature Communications — high-throughput robotic platform + active learning for electrolyte discovery
Co-authored work demonstrating an integrated robotic platform with active learning for accelerated discovery of optimal electrolyte formulations. The convergence of automation, ML, and battery chemistry.
Filed patent on selective ionic-species recovery membranes. Application of redox-flow membrane chemistry to lithium recovery.
After 14+ years at Argonne National Laboratory, started a new chapter in industrial data science, applied ML, and AI engineering.