I build data products, pipelines and climate risk analyses that turn scattered data into numbers you can trust, trace and act on.
Where events strike, and how hard.
storm tracks, flood depthWhat is in the way, and what it's worth.
buildings, assets, peopleHow much damage a given intensity causes.
damage curvesExpected annual loss, plus losses at 1-in-10 to 1-in-250-year events.
Every engagement starts with the data you already have. New tooling only comes in where it removes recurring manual work or answers a question you can't answer today.
Estimates of what floods, storms, heat and other hazards could cost your properties, operations or community, today and under future climate scenarios.
Raw data lands untouched, gets cleaned into one model you own, and every report is produced from that model rather than rebuilt by hand each month.
A fixed-scope review of one data flow, such as orders to invoices or inventory to reorders. You get a written map of where numbers come from, where they break, and what fixing each issue is worth.
CLIMADA (CLIMate ADAptation) is an open-source Python platform for probabilistic climate risk assessment, developed at ETH Zurich with a community of contributors. It combines hazard, exposure and vulnerability data to estimate the impacts of natural hazards worldwide.
Beyond today's risk, it can project impacts under climate change and socioeconomic scenarios, and compare the costs and benefits of adaptation options. Because the code and methods are public, every result can be inspected and reproduced.
I use CLIMADA to turn public hazard data and your own asset data into risk estimates you can plan around.
Curious Datum is independent and not affiliated with ETH Zurich or the CLIMADA project.
An interactive CLIMADA model is on the way. Pick a hazard, a place and a climate scenario, and see the expected impact.
This space is reserved for the live CLIMADA explorer. Until it launches, get in touch if you'd like a risk estimate for a specific place or portfolio.
Calculations are computed, never guessed. Running a report twice on the same data produces the same result.
A value nobody has mapped is held for a decision. Quietly filing it as "other" would make a gap look like an answer.
Raw data is kept separate from the rules applied to it. When a rule is corrected, past periods are rebuilt without re-collecting anything.
Where possible I use open-source tools like CLIMADA, so the method behind every number can be checked by anyone.
The data model, the mappings and the code are delivered to you. Nothing depends on a platform you can't leave.
Shops and local firms with years of sales, inventory or customer data spread across tools. I help connect it and turn it into a few numbers you check every week.
Property owners, local governments, nonprofits and businesses with physical sites who need to know what floods, storms or heat could cost, and which investments reduce it.
Load a product code like 002 with default settings and pandas reads it as the integer 2. Every lookup against it then misses, and no error is raised.
An expected annual loss can look small while a single 1-in-100-year event would be ruinous. Plans need both numbers.
Good hazard data is often public. What usually goes missing is an accurate list of what you own, where it is and what it's worth.
Coming next: worked case studies on public data, including CLIMADA runs on open storm-track and asset-value datasets.
A short email is enough: what you have, what you need to know, and where it hurts. I reply within two business days.