Northstar investigates your data like a senior analyst — profiling sources, building a plan, running analyses, verifying claims, and delivering grounded executive memos.
Unlike chat-with-data tools or notebook copilots, Northstar runs a structured investigation pipeline with built-in verification.
Creates a bounded analysis plan with risk flags before executing anything.
Trend analysis, segmentation, chart generation, and optional ML — all deterministic and reproducible.
Checks artifact coverage, evidence links, and claim quality. Surfaces caveats instead of hiding uncertainty.
Generates polished memo, technical appendix, and reproducible notebook — not just raw tables.
Ask follow-up questions and get answers grounded in the run's actual artifacts and findings.
Read-only, bounded retries, claim-to-evidence mapping, conservative language. Trust when it matters.
Every investigation follows a structured, auditable pipeline.
Register CSV, Excel, JSON, and text data sources.
Schema detection, quality signals, null rates, date columns, sample rows.
Infer entities, measures, dimensions, and candidate joins across sources.
Bounded analysis steps with risk flags. Optionally LLM-reasoned strategy.
Trend analysis, segmentation, chart generation, optional ML modeling.
Evidence audit, claim confidence scoring, caveat detection.
Executive memo, technical appendix, Jupyter notebook, grounded Q&A.
Existing tools force a tradeoff between autonomy and trust. Northstar delivers both.
| Capability | Chat w/ Data | Notebook Copilot | AutoML | Northstar ✦ |
|---|---|---|---|---|
| Multi-step investigation | Limited | Manual | No | ✓ 7-stage pipeline |
| Evidence-grounded claims | No | No | Partial | ✓ Every claim linked |
| Confidence levels | No | No | No | ✓ 3-tier rating |
| Built-in verification | No | No | No | ✓ Caveats + risk flags |
| Mixed-source reasoning | Limited | Manual | No | ✓ CSV + text + joins |
| Executive reports | No | No | Tables only | ✓ Memo + appendix |
Run your first investigation in under a minute.