Robert Rahardja · Chief Technology & AI Officer candidate
From satellite pixels to the number a bank discloses.
I build data platforms for regulated industries. This is a working miniature of a climate and nature-risk platform: open hazard and satellite data, scored for an illustrative APAC loan book, rolled up into IFRS S2 and TNFD disclosure metrics.
27%
Illustrative loan book · 2050 · high emissions
of SGD 945m in exposure is vulnerable to physical climate risk — the figure IFRS S2 ¶29(c) asks a bank to disclose.
See exactly how it is computed →- near protected nature
- 63%near protected nature
- in water-scarce basins
- 47%in water-scarce basins
- assets hotter by 2050
- 16/16assets hotter by 2050
The 30-second version
Builder first. Executive second.
The role is hands-on for 18 months, then leads a team. I have done both — in payments under MAS rules, in defence, in healthcare, and in my own companies.
Every requirement, mapped to evidence →10engineers
led to ship MAS TRM and PCI-DSS compliant payment systems on AWS
OxPay, 2023
LiveRAG co-pilot
in production for a paying client: own-document retrieval, multi-LLM routing
StarTech Innovation
50k+users
on a Ministry of Defence application I delivered and modernised
NCS Group, 2022–23
16years
building and running a medical-device company across Indonesia
2R Trade, 2002–18
One asset, end to end
How a pixel becomes a disclosure line.
Industrial estate, Ayutthaya, Thailand. Industrial estates here were under water for weeks in the 2011 Chao Phraya floods. Four steps, every value real and traceable.
01
Read the hazard cube
45 flood-depth layers
WRI Aqueduct river-flood depths — 5 climate models × 9 return periods — read straight from a cloud Zarr store. No downloads, no copies.
02
Locate and sample
14.350°N, 100.580°E
The asset’s ~1 km cell, with neighbours checked so a dry pixel in a flooded city is never reported as “safe”.
03
Compute the metric
0.20 → 0.72 m
1-in-100-year depth today versus 2050 under RCP8.5; 0.60–0.91 m across models.
04
Roll up to disclosure
SGD 110m
Above the 0.5 m threshold, so its exposure counts toward IFRS S2 ¶29(c) — with the definition written down.
The climate signal
Every asset gets hotter. Some, dramatically.
Days a year with a daily mean above 30 °C — the median of six downscaled CMIP6 models, today versus 2050 under a high-emissions scenario. Bangkok and Ayutthaya cross 250 days. Western Sydney barely moves. Same portfolio, very different risk.
Per-model values and ranges are on each asset card in the demo — the spread between models is information, not noise.
See it work
The live demo
16 assets in 7 countries on a map, a risk card per asset, and the disclosure view a bank would publish.
Open →
Think it through
How I’d build it
Architecture, the first 90 days, an 18-month roadmap, first hires, partners, and the governance banks will ask for.
Open →
Look ahead
Tech roadmap
The stack at three stages — funded startup, scaling, enterprise-grade — what triggers each move, and what never changes.
Open →
Check the work
Methodology & data
Every source, licence, threshold and limitation. A climate number without its caveats is not decision-useful.
Open →