01
Import the messy sheet
Bring in CSV, Excel, or project tables from your local workspace. IMD starts from the raw file you already have, not from a synthetic demo dataset.
AI Data Cleaning, Not Another AI Dashboard
Turn raw datasets intoanalysis-ready results.
IMD is a local-first desktop workspace for data cleaning. Instead of asking the user to drive every operation, it lets AI organize the preparation workflow and hands the user a result to review, confirm, and export.
Built for analysts, researchers, students, and teachersLocal workspace + desktop shell + CLI runtime
Result Package
DID-ready panel dataset
Local review state
System found
- Duplicate `(city_id, year)` rows removed
- Treatment labels normalized to binary flags
- Missing policy-year records isolated for review
Output
`province_panel_did_ready`
New clean table created with snapshot preserved.
User role
Review and confirm
The system performs the prep work. The user approves the result.
Workflow
A service-shaped interface for data preparation
The system should not merely help people click faster. It should take responsibility for the repetitive, low-value work of turning messy data into something a serious downstream tool can trust.
02
Let the system prepare the result
IMD diagnoses structure, cleans values, standardizes fields, and prepares analysis-ready tables through a local AI runtime and internal capabilities.
03
Review, confirm, export
You stay in control as the reviewer. Accept the cleaned result, inspect changes, and move downstream into Stata, R, or your reporting workflow.
Capabilities
Structured like a desktop product, not a chatbot wrapped around a sheet
AI-led delivery
The user sets the goal. IMD organizes the cleaning workflow, calls internal capabilities, and returns a result package instead of asking the user to assemble steps by hand.
Local-first trust
The desktop shell and local CLI runtime keep the working loop close to the user, which is critical for sensitive research files and institutional datasets.
Built for dirty data
Missing values, inconsistent category labels, duplicate keys, broken date columns, and panel-data preparation are treated as the product core, not as edge cases.
Project-based workflow
Projects, tables, versions, and snapshots stay organized as real working objects. IMD is not just a chatbot wrapped around a spreadsheet.
Product Positioning
IMD sits upstream of serious analysis software
The goal is not to replace Stata, R, or Tableau. The goal is to remove the ugliest part of the work before those tools become useful: cleaning, standardizing, shaping, and preparing data into trustworthy structures.
Not another analytics copilot. IMD is for the painful data-prep layer before analysis starts.
Not tool-first. The system should deliver a usable result; the user should primarily review and confirm.
Not all-purpose. The wedge is narrow and deep: cleaning, structuring, panel preparation, and export readiness.
Desktop Product
A calmer front-end for the noisiest part of analysis work
IMD is designed for people who trust their conclusions only when they trust the data underneath. That makes data cleaning the product core, not a side utility.
View the project
Local workspace. Review-first workflow. Export-ready output.