About Kaiso
Making AI-powered data work auditable, verifiable, and human-scale.
Where Kaiso began
Kaiso was born from a specific, frustrating reality: AI tools struggle with complex, real-world data work. The idea crystallized during a consulting engagement with an investigative data journalism team at Grist, who were working with niche formats — GeoJSON, land records, spatial datasets — for their investigation into land-grant universities and stolen Indigenous land. Building a custom data engine for that project revealed just how bespoke and demanding this kind of analysis can be.
When the team later adopted AI tools for related work, new challenges emerged. Outputs were difficult to audit. Verification happened only at the end instead of incrementally. Documenting methodology became its own burden. The tools were powerful in isolation, but they didn't fit the way careful, accountable data work actually happens.
The vision
Kaiso brings LLM-powered data transformation to human scale. Instead of dumping everything into a chat window and hoping the output is right, Kaiso enables a different workflow: continual, gradual auditing and verification of correctness and intent. You work in smaller, individually iterable chunks — each one inspectable, each one correctable — with automated methodology reporting built in.
Describe your transformations in plain language and Kaiso generates Python pipelines you can review and refine. Your data never leaves your machine. Your workflows are shareable and reproducible. The result is AI that works the way practitioners actually need it to — transparent, incremental, and accountable.
Team
After nearly fifteen years building machine-learning and data-intensive systems, one pattern kept repeating: the gap between what AI can do in a demo and what it can do for someone working with real data is enormous. Making AI genuinely useful means meeting practitioners where they are, not asking them to reshape their work around the tool.
I saw firsthand how AI tools fail real practitioners: outputs that can't be audited, workflows that can't be reproduced, and a "trust me" approach that's fundamentally at odds with rigorous data work. Kaiso is the tool I kept wishing existed — one that treats verification and transparency not as afterthoughts, but as core to the experience.