Skip to content
View Rekin226's full-sized avatar

Highlights

  • Pro

Block or report Rekin226

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Rekin226/README.md

Abdoul Rachid Ouédraogo (韋正)

Typing SVG

Applied AI Researcher @ Taiwan Polar Institute · Adjunct Instructor @ FCU · Founder of POUK YAM

Google Scholar LinkedIn Website Location


I build AI for science: physics-informed machine learning, agentic AI, and the tested software that ships it. My depth is in hydrogeology and groundwater modeling, which I use as a proving ground for honest, reproducible scientific ML. I also teach numerical methods at Feng Chia University.


🚀 Featured projects

PyTorch CUDA

GPU physics-informed neural operators for groundwater, one model across 61 wells.

  • One attribute-conditioned operator vs 61 hand-calibrated ODEs
  • Generalizes to unseen wells: leave-one-well-out KGE 0.565 (climatology 0.446, in-sample 0.591)
  • Forecast LSTM beats persistence at 7 & 30 days, calibrated intervals
  • 14× GPU speedup (CUDA, bf16) · NVIDIA PhysicsNeMo port

Live demo

Python PyPI

Open-source Python toolkit that unifies global water data under one API.

  • 29 collectors behind one schema: USGS, UK EA, Hub'Eau, GRDC, BOM, FAO, GEMStat, EU WFD
  • 26 methodologies scored and auto-executed by an AI engine
  • MCP server + aquascope ask — Claude or Cursor gets stations, series, and cited flood frequency
  • Bulletin 17C · FAO-56 · CAMELS-validated · 1,478 tests · 15 outside contributors

Explorer Support on Ko-fi

Python LangChain

Tool-calling aquaponics agent over a deterministic, tested, cited engineering core.

  • LLM collects and routes; a verifiable trust zone does the math
  • Photos and voice notes on Telegram, WhatsApp, and the web
  • Vision only observes: a guard strips fabricated readings, and a cited table returns a ranked differential, never a verdict
  • 541 tests · advice-safety golden set enforced in CI · runs on self-hosted open weights

Claude Code plugin Semantic Scholar

Claude Code plugin that turns the agent into a disciplined manuscript collaborator.

  • Five modes: draft, review, revise, proofread, audit
  • Semantic Scholar MCP citations (no API key), anti-fabrication guardrails
  • Hydrology (HJ, JHRS) + IEEE + a generic quantitative-science profile
  • Clean .docx round-trip; bundled try-it demos

🔬 Research interests

Physics-informed & scientific ML · Agentic AI for science · Groundwater modeling & hydrogeology · Numerical methods & scientific computing · Sustainable water resources

🛠️ Tech stack

tech stack


📌 Recent work

  • 🌊 AquaScope v0.10.0: 29 data collectors under one schema, an MCP server that hands Claude and Cursor real gauge data, and the Explorer — 45,919 public gauges on one map, with flood frequency computed in the browser on Pyodide, no server. The open archive behind it, Rekin226/aquascope-gauges, mirrors discharge, water level, rainfall and borehole groundwater as weekly-harvested GeoParquet
  • 🌱 Agronaut: the aquaponics agent gained field senses — photos and voice notes on Telegram, WhatsApp and the web. A vision model only observes; a deterministic guard strips any fabricated reading or prescription out of its description, and a cited triage table returns a ranked differential with the checks that discriminate, because a photograph cannot separate iron deficiency from pH lockout. 541 tests, an advice-safety golden set that fails the build on a regression, and a prepared Digital Public Good compliance pack (privacy, AI transparency, SDG mapping)
  • 🧠 HydroPhysicsAI: GPU physics-informed neural operators benchmarked against per-well gray-box ODEs, with a live demo and an honest, reproducible benchmark harness
  • 🤖 paper-agent: packaged as a Claude Code plugin (v1.5.0) — draft/review/revise/proofread/audit manuscripts with Semantic Scholar MCP citation integrity, now broadened from hydrology to IEEE and a generic quantitative-science profile
  • 📄 Hydrogeology Journal accepted manuscript on gray-box groundwater modeling for the Zhuoshui Alluvial Fan, Taiwan
  • 📚 Publications & patent: 4 peer-reviewed papers (3 SCIE-indexed, incl. Hydrogeology Journal 2026) · granted Taiwan patent M661364
  • 🎓 Numerical Analysis with Python, graduate course taught at Feng Chia University
  • 💼 POUK_YAM, consulting and software at the intersection of scientific computing and agentic AI

🎓 Background

  • Ph.D., Infrastructure Planning & Engineering (groundwater modeling & hydrogeology), Feng Chia University, Taiwan · 2023
  • M.S., Water Resources Engineering & Conservation, Feng Chia University, Taiwan · 2019
  • Undergraduate, 2iE (Institut International d'Ingénierie de l'Eau et de l'Environnement), Ouagadougou, Burkina Faso

🌐 Multilingual: French (native) · English (advanced) · Mandarin (working proficiency)


📈 GitHub activity

stats top langs streak trophies

POUK_YAM AquaScope

Building AI for science · Taichung, Taiwan

Pinned Loading

  1. aquascope aquascope Public

    Open-source Python toolkit for water data, hydrology, and agricultural water management — 29 unified collectors (USGS, UK EA, Hub'Eau, GRDC, BOM, FAO…), Bulletin 17C flood frequency, FAO-56 ET₀, an…

    Python 19 22

  2. paper-agent paper-agent Public

    Claude Code plugin to draft, review, revise, proofread, and audit quantitative-science journal manuscripts at submission quality. Five strict modes, Semantic Scholar MCP citation integrity (no API …

    8 2

  3. HydroPhysicsAI HydroPhysicsAI Public

    GPU physics-informed neural operators for groundwater, benchmarked against per-well gray-box ODEs (PyTorch / PhysicsNeMo / CUDA)

    Python 3

  4. Agronaut Agronaut Public

    An open-source AI agronomy agent that runs locally on your own machine or server, reaches you on Telegram and WhatsApp, and computes aquaponics system designs from a deterministic, source-cited eng…

    Python 1 4

  5. aquascope-demos aquascope-demos Public

    Real-world hydrology demo cases for AquaScope. Self-contained reproducible Jupyter notebooks with every headline result checked against a published reference.

    Jupyter Notebook

  6. groundwater-prediction-graybox-model groundwater-prediction-graybox-model Public

    Gray-box modeling workflow for groundwater level prediction in the Zhuoshui Alluvial Fan, Taiwan. Companion code to a Hydrogeology Journal manuscript.

    Python