From Personas to Simulated Users: A Fitness-for-Purpose Survey
Liu, X., et al. (including Wu, K.) (2027). "From Personas to Simulated Users: A Fitness-for-Purpose Survey." Submitted to the AAAI 2027 Artificial Intelligence for Social Impact Track.
CV - Koutian Wu | AI4Geoscience M.S. Student at UT Austin
Highlights
Received Tencent Qingyun Talent Program: Hunyuan AI Data & Evaluation Offer Aug. 2026 - Open-source contributor to repositories totaling 100K+ stars (git/git, multica, ai-agent-book, AionUi, harbor); 10K+ GitHub contributions in the last year
- Open to research internships in AI Agent Development/Test/Evaluation, ML/DL/DS, and Full-Stack Engineering
- 2+ years work experience (US/International); 4+ years research/programming experience
- Invited reviewer for Planetary and Space Sciences (Elsevier, Scopus Q2), with further review invitations from IEEE TNNLS, ACM Computing Surveys, ACM Transactions on Sensor Networks, the COLM workshop track, and ACM AgentSkills
Education
- M.S. Student in AI for Geosciences, University of Texas at Austin, 2024 - Expected: Dec 2026 (transferred from the Ph.D. track)
- Graduate Teaching Assistant in Fall 2024, Spring 2025, Spring 2026, and Fall 2026
- Advisors: Dr. Zong-Liang Yang
- B.S. with Honors in Space Physics (Honors Class), University of Science and Technology of China, 2020 - 2024
- Nominated for the Highest Honor for USTC undergraduates (one of the two ESS School nominees)
- Founded one of the largest student clubs, grew from 0 to 1,300+ members. News
Work Experience
- Core Founding Team Member (Jan 2026 - Jul 2026)
- Starting Point One (SP1), Remote
- Sourced 100+ talents for Kehan Dong (CEO of SP1, ex-YC China Partner); sent 1,000+ outreach emails supporting post-investment hiring for Mira (raised $6.6M from General Catalyst)
- Graduate Teaching Assistant (Jan 2026 - Present)
- The University of Texas at Austin
- TA for GEO 302C in Spring 2026; current Fall 2026 appointment
- Graduate Research Assistant (Aug 2025 - Dec 2025)
- Jackson School of Geosciences, University of Texas at Austin
- AI for Noah-MP Land Surface Model: implementing rock and wood moisture into the Noah-MP land surface model
- Project Lead, High-Performance Computing Allocation (Aug 2025 - Aug 2026)
- NSF NCAR, Remote
- Proposed and secured NSF NCAR allocation: 1k GPU + 22k CPU hours.
- Integrating machine learning parameter calibration for a SOTA physics-based land surface model (Noah-MP).
- Developing a multi-expert AI agent system for automated parameterization of physics-based climate models.
- Lead Developer, ESM-bench (Dec 2025 - Present)
- UT Austin
- Designed a 243-task benchmark and multi-model evaluation harness measuring where frontier models fail on real scientific code. Preprint
- AI Intern, AI Agent Knowledge Base Evaluation (Jul 2025 - Aug 2025)
- 19Pine.AI (Singapore), Remote
- Implemented a multi-dimensional evaluation system from scratch for PineAgent’s RAG knowledge base.
- Developed a data sanitization module removing 2,000+ PII entries and noise from 1,000+ call sessions.
- Extracted 3,000+ Q&A pairs (knowledge/method/strategy) from call sessions to form the evaluation dataset.
- Deployed concurrent LLM-as-judge + rule-based engine, measuring precision/recall/F1 to enable RLHF optimization.
- Full-Stack Intern, LLM Text Processing System (Jun 2025 - Jul 2025)
- ZaiwenAI.com, Beijing, China
- Developed a 3-module MVP: LLM-footprint detection, removal, and plagiarism checking for researchers.
- Built a RESTful backend with FastAPI and an asynchronous task queue with Celery + Redis.
- Created a Vue.js frontend with 9-format document upload and SSE-based real-time LLM response streaming.
- AI Consulting Expert (May 2025 - Jun 2025)
- Invited by the CEO, Keshi Edu, Remote (Beijing)
- Designed an architecture for a RAG-based knowledge base to customize student graduate program applications.
- Identified critical bugs and exposed API keys, helping avoid significant potential losses.
- Graduate Teaching Assistant (Aug 2024 - May 2025)
- The University of Texas at Austin
- Supported 1,013 students across two semesters of Earth in 2100 (GEO 303E)
- Earned seven-item course-evaluation means of 4.09/5 in Fall 2024 and 4.49/5 in Spring 2025
- Scientific Visitor (Jun 2024 - Aug 2024)
- Peking University, Beijing, China
- Hydrology and Water Resources Science, model evaluation of physics-based land surface models; Advisor: Prof. Peirong Lin
- Visiting Scholar (Jul 2023 - Dec 2023)
- NSF NCAR, Boulder, Colorado
- Research on volcanic perturbations in the MLT region using WACCM-X simulation and meteor radar observations
- Re-confirmed that a single volcano can send shockwaves around Earth, seen in simulations and observations
- Student Fellow (Oct 2021 - Feb 2024)
- Shenzhen X-Institute (A Shenzhen Government-Tsinghua Univ. Initiative)
Research Interests
- AI Agent Evaluation for Science: benchmarks and harnesses for whether AI agents understand scientific models and code
- AI for Land Surface Modeling: implementing explainable AI/ML to improve physics-based Noah-MP land surface modeling
- High Performance Computing: GPU and CPU clusters for large-scale geoscience simulations
- Model Optimization: physics-based land surface models including Noah-MP, CTSM, HRLDAS
Skills
- Programming: Python, JavaScript/TypeScript, MATLAB, Fortran, Shell, SQL
- AI/ML: AI Agent Evaluation, LLM-as-a-judge, benchmark design, LLM APIs, Prompt Engineering, PyTorch, LangChain, RAG, HuggingFace
- Frameworks and Tools: FastAPI, Celery, Redis, Vue.js, React, Docker, Git, HPC, AWS/GCP
- Geoscience: Land Surface Model Development (Noah-MP, CTSM, HRLDAS), WACCM, Model Evaluation
Publications
ASI-Bench: At the Dawn of Artificial Superintelligence
Zhou, J., et al. (including Wu, K.) (2026). "ASI-Bench: At the Dawn of Artificial Superintelligence." arXiv preprint arXiv:2608.17271.
MicroVerse: An Instrument for Measuring Self-Authored Identity Drift in Long-Horizon Multi-Agent Language-Model Simulations
Ng, S., et al. (including Wu, K.) (2026). "MicroVerse: An Instrument for Measuring Self-Authored Identity Drift in Long-Horizon Multi-Agent Language-Model Simulations." arXiv preprint arXiv:2608.15844.
MatrAIx: Simulating the World with 8.3 Billion Persona Agents
Li, X., et al. (including Wu, K.) (2026). "MatrAIx: Simulating the World with 8.3 Billion Persona Agents." arXiv preprint arXiv:2608.04205.
ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research
Xu, W., et al. (including Wu, K.) (2026). "ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research." arXiv preprint. Submitted to AAAI 2027.
How Does Integrating Plant Hydraulics Improve Noah-MP Land Surface Model
Wu, K., Li, L., Rempe, D., Matheny, A., Mbarak, M., & Yang, Z.-L. (2026). "How Does Integrating Plant Hydraulics Improve Noah-MP Land Surface Model." Poster presented at the 106th AMS Annual Meeting, Houston, TX.
On the Ethics of Generative GeoAI: Explainability, Bias, Hallucination, Accountability, Privacy, and Trust
Mai, G., Lao, N., Zhang, J., Mao, L., Wang, Z., Wu, N., Janowicz, K., Wu, K., Rao, J., Gao, S., & Zhu, R. (2026). "On the Ethics of Generative GeoAI: Explainability, Bias, Hallucination, Accountability, Privacy, and Trust." In Geography According to Foundation Models, Vol. 422, pp. 215-232. IOS Press. DOI 10.3233/FAIA260483.
ESM-bench: A Benchmark for Evaluating Whether AI Agents Understand Earth System Model Physics and Code
Wu, K., Cao, Y., & Mai, G. (2026). "ESM-bench: A Benchmark for Evaluating Whether AI Agents Understand Earth System Model Physics and Code." Preprint, Zenodo. https://zenodo.org/records/19802836
Noah-Agent: A Multi-Expert AI Agent Framework for Automated Parameterization and Validation of Large-Scale Fortran Climate Models (v0.1)
Wu, K. (2025). "Noah-Agent: A Multi-Expert AI Agent Framework for Automated Parameterization and Validation of Large-Scale Fortran Climate Models (v0.1)." Preprint, Zenodo. https://zenodo.org/records/17862049
Diurnal and seasonal variations of meteor speed and arrival angle observed by Mengcheng meteor radar
Wu, K., Yi, W.*, Xue, X.*, Reid, I., & Lu, M. (2024). "Diurnal and seasonal variations of meteor speed and arrival angle observed by Mengcheng meteor radar." JGR: Space Physics.
A Summary Report on the Space Physics Practical Education in 2022
Wu, K.*, Xu, X., Jiang, J., & Shen, A. (2024). "A Summary Report on the Space Physics Practical Education in 2022." Review of Geophysics and Planetary Physics.
Perturbations by the 2022 Hunga-Tonga Volcano Eruption in the MLT Region Investigated Using the WACCM-X Simulation and Meteor Radar Observations
Wu, K., Liu, H.-L., Yi, W., & Xue, X. (2023). "Perturbations by the 2022 Hunga-Tonga Volcano Eruption in the MLT Region Investigated Using the WACCM-X Simulation and Meteor Radar Observations." AGU Fall Meeting Abstracts, SA33B-2892.
Talks
ESM-bench: Can AI Agents Understand Earth System Model Physics and Code?
Oral Presentation at 4th Annual Good Systems Smart Cities and AI Innovations Symposium, Austin, TX
Noah-MP land surface model with plant hydraulics scheme (Noah-MP-PHS) Evaluation
Poster at 106th AMS Annual Meeting, Houston, TX
An open-source geospatial AI platform based on the AlphaEarth Foundation Model
Oral Presentation at Geoscience Hackathon '25, Austin, TX
Noah-MP land surface model with plant hydraulics scheme (Noah-MP-PHS) Evaluation
Poster at Advancing Land Modeling Symposium, Austin, TX
Noah-MP land surface model with plant hydraulics scheme (Noah-MP-PHS) Evaluation
Poster at JSG Research Symposium, Austin, TX
Perturbations by the 2022 Hunga-Tonga Volcano Eruption in the MLT Region Investigated Using the WACCM-X Simulation and Meteor Radar Observations
Poster at AGU Fall Meeting 2023, San Francisco, CA
Perturbations by the 2022 Hunga-Tonga Volcano Eruption in the MLT Region Investigated Using WACCM-X Simulation and Meteor Radar Observations
Oral Presentation at 4th NASA DRIVE CGS Workshop, Boulder, CO
Teaching
Service and Leadership
- Project Admin, NCAR/CISL Allocation Project UTAA0012 (Aug 2025 - Present)
- Technical leadership and implementation of project workflows for “Explainable AI for Improving Physics-Based Noah-MP Land Surface Modeling”
- President & Founder, USTC Student Xingyun Poetry Club (Nov 2020 - Jun 2022)
- Founded USTC’s first poetry club, grew to 1,300+ members
- Organized 25+ cultural events with 202+ attendees
- Managed team of 8 co-founders and 4 vice presidents Research Projects ======
- Perturbations by the 2022 Hunga-Tonga Volcano Eruption in the MLT Region
- Investigated using WACCM-X Simulation and Meteor Radar Observations
- Abstract: https://ui.adsabs.harvard.edu/abs/2023AGUFMSA33B2892W/abstract
- Re-confirmed that a single volcano can send shockwaves around Earth
Awards and Honors
- NSF NCAR Computational Resource Allocation (Aug 2025)
- 1,000 GPU hours on NSF NCAR Derecho-GPU
- 20,000 CPU core-hours on NSF NCAR Derecho
- 2,000 CPU core-hours on Casper
- 2 TB campaign storage