Drop-in CSV + Excel
Drag a file (CSV, XLSX, TSV) into the workspace. Aixys infers column types, parses dates, detects encoding, and offers an interactive preview within seconds.
Pillar VI · Dataset Workspace
Where the dataset becomes the analysis.
The Dataset Workspace is the room where research data is unpacked. Drop a CSV or Excel file and Aixys profiles every column, flags issues, surfaces the natural cuts, and drafts an interpretation you can refine. From there you launch any of the other pillars — design, statistics, predictive — against the same dataset with the lineage preserved.
Workspace · drop file
CSV · XLSX · TSV
Drop a CSV, XLSX, or TSV here
— or browse files —
UP TO 250 MB·UTF-8 / 16·EN-CA · EN-US · DE
01 · drop & profile
The first three seconds tell the story
Drop a file and Aixys profiles every column, infers types, parses dates, and produces a preview you can act on — before the loading animation has finished.
| plot | block | yield_t_ha | n_rate | rain_mm | soil_n_kg | cultivar | obs_date |
|---|---|---|---|---|---|---|---|
| A-01 | A | 5.42 | N1 | 342 | 38 | C1 | 2026-03-12 |
| A-02 | A | 6.18 | N2 | 342 | 41 | C2 | 2026-03-12 |
| A-03 | A | 4.92 | N0 | 342 | — | C3 | 2026-03-12 |
| A-04 | A | 6.81 | N3 | 342 | 44 | C2 | 2026-03-12 |
| B-01 | B | 5.71 | N2 | 318 | 37 | C1 | 2026-03-12 |
| B-02 | B | 4.36 | N0 | 318 | 32 | C3 | 2026-03-12 |
| B-03 | B | 6.42 | N3 | 318 | — | C2 | 2026-03-12 |
| B-04 | B | 5.89 | N1 | 318 | 36 | C1 | 2026-03-12 |
| — 3,392 more rows hidden — | |||||||
| Column | Type | Distribution | Missing |
|---|---|---|---|
| yield_t_ha | numeric · float | μ 5.61 · σ 0.92 · range 2.81–7.94 | 3% |
| n_rate | categorical · 4 levels | N0 (25%) · N1 (25%) · N2 (25%) · N3 (25%) | 0% |
| soil_n_kg | numeric · float | μ 36.4 · σ 4.8 · range 22–48 | 18% |
| cultivar | categorical · 3 levels | C1 (33%) · C2 (34%) · C3 (33%) | 0% |
| obs_date | date · ISO 8601 | 2026-03-12 → 2026-09-04 | 4% |
02 · interpret
The dataset, in plain English
A written interpretation generated from the actual structure of the data — what it appears to measure, how it is shaped, where the natural cuts live, what questions it can credibly answer.
AI summary · drafted from dataset
The dataset describes a spring wheat field trial across two sites (Hayfield, Brookway), measuring yield (t/ha) by cultivar (3 levels) × nitrogen rate (4 levels) within 4 randomised complete blocks per site. Date range spans March 2026 to September 2026, with weekly observation cadence.
The structure is consistent with a factorial RCBD. The natural cuts are cultivar, nitrogen rate, and site; the natural response is yield. 18% of soil-N values are missing, concentrated in Block C — this should be triaged before any N-sensitive model.
Three analyses are ranked by fit below. The dataset is analysis-ready at 86 / 100 — see the data-quality report for the punch list of fixes.
03 · launch
Ranked analyses, one click away
The workspace recommends the analyses that fit the dataset, ranked by goodness-of-fit and statistical power. Each card carries the reason for the suggestion and the path into the right pillar.
01 · Two-way ANOVA
Yield ~ Cultivar × N rate · Block
Factorial structure detected (3 cultivars × 4 N rates, balanced within 4 blocks). Residual variance permits power 0.89 to detect Δ = 0.35 t/ha.
Launch in workspace →
02 · Field-trial layout
Visualise as split-plot map
Block + plot columns recognised as a Randomised Complete Block Design — open in the Experimental Design Assistant for randomisation audit.
Launch in workspace →
03 · Yield forecast
Project end-of-season yield by field
Time-indexed observations with three covariates (rain, soil N, cultivar) are sufficient for a gradient-boosted model. Out-of-time holdout available.
Launch in workspace →
04 · capability
What ships in the workspace
Six core capabilities that turn a raw file into something you can analyse, share, and report.
AI-drafted summary
A written narrative of the dataset — what it appears to measure, how it is structured, where the obvious cuts and segments live, and what questions it can credibly answer.
Inline validation
Every column inspected for missingness, outliers, unit drift, encoding errors, and impossible values. Issues are listed with severity, count, and a recommended next action.
Chart previews
Distribution, time-series, correlation, and group-comparison previews appear automatically. Click any preview to promote it into a saved figure or pass it to the report generator.
AI recommendations
Suggested analyses ranked by fit to the dataset — "this looks like a factorial trial, consider two-way ANOVA on Yield by Cultivar × Site." Each recommendation links to the right pillar.
Export insights
Send any view, figure, or summary into the Report Generator. Snapshot the workspace to a shareable read-only URL or export the cleaned dataset back as CSV.
05 · in practice
What teams use it for
Three representative workflows drawn from agronomy, lab QC, and operations.
Agronomist drops a 3,400-row trial dataset for a quick scan
Workspace shows a balanced split-plot, flags one suspect block (yield 38% below site mean), and suggests an ANOVA route — analysis launched with one click.
Chemist uploads a 96-well plate export from an assay reader
Plate map rendered with control wells highlighted, two outlier wells flagged with z-score, and a 4PL dose-response preview offered.
Process engineer uploads a month of throughput logs
Time-series preview surfaces a weekly seasonal pattern; the workspace recommends an exponential smoothing forecast and a regression against three sensor inputs.
06 · sectors
Where this belongs
Industries with the heaviest dataset cadence. Templates accelerate the work; the engine is general.
AgricultureBiotechLab QCField researchOperationsConsultingAcademic groups
Continue · 06
Bring a real dataset to the workspace.
A 30-minute working session with the Aixys studio. Bring a CSV or Excel file from your team's current research; leave with an interpretation, a ranked analysis plan, and a reusable workspace.