What we do
Four ways to read your data.
Each one starts from a question someone in your business needs answered, and ends with something they can read without us.
§ 1
Analytics
Questions about your business, answered from your own records.
We start from the question you need answered, find the data that can answer it, and write up what it shows and how sure we are. Every finding comes with the query behind it, so you can check it.
- Questions turned into queries
- Findings written in plain language
- The working shown, so you can check it
§ 2
Dashboards
The few figures your team should see every week, kept honest.
We pick the handful of measures that change decisions, define each one in writing, and build the dashboard in the tool you already use or as a small web app of its own.
- Measures defined in writing
- Built in Metabase, Looker Studio or Power BI
- Or a dashboard of its own in React
§ 3
Data pipelines
Data moved from where it is made to where it can be read.
We pull data out of your apps, payment systems and spreadsheets into one warehouse, model it into tables people can query, and schedule the loads so the figures stay fresh and tested.
- Extraction from apps, APIs and spreadsheets
- Modelled tables with tests
- Scheduled, monitored loads
§ 4
AI over your own data
Search and answers over your documents and records, with sources.
We build assistants that answer questions from your own files and databases and show where each answer came from. We test them on questions your documents cannot answer, so you see how they behave when they do not know. They run under your own cloud and AI provider accounts, so you choose where your data goes.
- Document search and question answering
- Answers that cite their sources
- Built in your own cloud and AI accounts
Tbl. 1, the tools we work with
| Warehouse and storage | PostgreSQL, BigQuery, Snowflake |
|---|---|
| Pipelines | Python, SQL, dbt, Airflow |
| Dashboards | Metabase, Looker Studio, Power BI, React |
| AI over your data | OpenAI and Anthropic APIs, LangChain, pgvector |
We work remotely, on US time zones, in English.