Skip to content

Researcher Profile

The ResearcherProfile is a persistent record of who you are as a researcher — your expertise, preferred models, active projects, and domain focus. It is built up from agent-researcher interactions over time and injected into every conversation automatically.

What It Stores

~/.aihydro/researcher.json
{
  "name": "Mohammad Galib",
  "institution": "Purdue University",
  "role": "PhD Researcher",
  "domain": "Computational Hydrology",
  "expertise": ["watershed modelling", "differentiable hydrology", "CAMELS benchmark"],
  "tools_familiarity": {
    "HBV-light": "advanced",
    "NeuralHydrology": "intermediate",
    "PyTorch": "intermediate"
  },
  "preferred_models": ["HBV-light", "LSTM"],
  "research_focus": "Investigating the role of geology in controlling baseflow generation across CAMELS-US catchments.",
  "active_project": "New England Basins",
  "communication_style": "concise, technical",
  "observations": [
    "Prefers NSE and KGE together rather than NSE alone for model evaluation.",
    "Tends to work with 20-year streamflow records for signature extraction.",
    "Interested in spatial patterns more than single-basin deep dives."
  ]
}

How It's Built

The profile is not filled in manually (though you can). It accumulates automatically:

  • When you correct the agent, it logs the observation
  • When you consistently use certain tools or parameters, it notes the pattern
  • When you start a project or focus on a domain, it updates your active context

The agent calls log_researcher_observation silently when it learns something meaningful about your preferences.

How It's Used

At the start of every conversation, the agent calls get_researcher_profile() and uses the result to:

  • Skip beginner explanations if your expertise is advanced
  • Default to your preferred model when you say "calibrate a model" without specifying which
  • Use your preferred metrics when reporting results
  • Reference your active project when starting work without explicit context
  • Tailor the communication style — concise and technical vs exploratory and explanatory

This is the same idea as the memory features in Claude.ai and ChatGPT, but domain-specific to computational hydrology.

Managing Your Profile

View your profile

Show me my researcher profile.

Update a field

Update my research focus to: investigating the role of snow in modulating
streamflow seasonality across the Pacific Northwest.

Set my active project to "Pacific Northwest Basins".

Add to expertise

Add "snow hydrology" to my expertise areas.

Reset observations

Clear my researcher observations — I want a fresh start.

Profile in research.md

The profile is also appended to .aihydrorules/research.md in your workspace on every session save — so even if an agent doesn't call get_researcher_profile, the persona context is injected automatically via the rules file.

.aihydrorules/research.md (excerpt)
## Researcher Profile
- **Name:** Mohammad Galib
- **Role:** PhD Researcher — Purdue University
- **Domain:** Computational Hydrology
- **Expertise:** watershed modelling, differentiable hydrology, CAMELS benchmark
- **Active Project:** New England Basins
- **Preferred Models:** HBV-light, LSTM
- **Communication style:** concise, technical