Technology choice depends on context
Flood exposure, cyclone risk, salinity and hydro-geology all change which WASH technology actually fits — reasoning that wasn't available in one place to query.
Illustration · representative concept, not a live system screenCase study · DPHE-WASH · AI budget assistant
Turning a national WASH technology catalogue into instant, grounded budget guidance.
Tiger Park built and operates a public compendium of water, sanitation and hygiene technologies for Bangladesh's disaster-response sector — complete with real per-technology costs, a live project map, and an AI assistant that turns a plain-language emergency scenario into an itemized budget grounded in that same cost data.
ScopeWater, sanitation and hygiene technologies, for disaster response nationwide
Initiative contextNational Compendium of WASH Technologies for Disaster Response — DPHE, ITN-BUET, UNICEF, Bangladesh WASH Cluster
Built and operated byTiger Park Limited
15 locationsWASH projects plotted live on the public map
6 categorieswater, sanitation & hygiene, each for during- and post-disaster response
GPT-OSS 120Bthe model behind Naya, the grounded budget assistant
EN + BNthe full public interface, bilingual by default
The assignment
Choosing the right WASH technology for a disaster meant knowing a lot, and finding it fast.
Which water, sanitation or hygiene technology fits a given emergency depends on flood exposure, salinity, arsenic risk and how deep the water table sits — and that judgment historically lived across specialists' heads and long reference documents, not somewhere a field planner could query directly. Budgeting compounded the problem: comparing per-unit costs across dozens of technology options for a specific household count and scenario was a manual, spreadsheet-style exercise, redone from scratch each time.
DPHE-WASH · public technology referenceTiger Park built and operates the compendium and its AI assistant.
DPHE hosts the public compendium at its own subdomain, and the initiative sits alongside ITN-BUET, UNICEF and the Bangladesh WASH Cluster's work in the sector. Tiger Park designed, built and runs the system, including Naya, the AI budget assistant.
Challenge
Turn scattered technical knowledge into one browsable, costed reference.
Budgeting meant manual cross-referencing
Comparing per-unit costs across technology options for a specific household count and scenario was redone by hand each time, from static references.
No shared reference existed
DPHE, ITN-BUET, UNICEF and the Bangladesh WASH Cluster each held pieces of this knowledge, with no single public interface bringing technology, cost and location together.
The system
A live public compendium, mapped projects, and an assistant that does the arithmetic.
The public site brings the technology catalogue, real cost data, a project map and the Naya assistant together in one bilingual interface, built and hosted by Tiger Park.
A structured technology catalogue
Water supply, sanitation and hygiene technologies, each classified for during-disaster or post-disaster response, with technical descriptions and design diagrams.
Real unit costs in BDT
Every technology carries a costed estimate drawn from the underlying database — from a ৳52,123 dual-platform tubewell to a ৳37.8 lakh desalination system.
An interactive project map
WASH project locations plotted across Bangladesh's divisions, filterable by water, sanitation or hygiene, searchable by place name.
Naya, a grounded budget assistant
A chat assistant, running on GPT-OSS 120B, that takes a plain-language emergency scenario and returns technology recommendations, engineering rationale and an itemized, multi-scenario budget — grounded in the same cost database, not invented on the fly.
Bilingual by default
The full public interface is available in English and Bangla.
Built for a multi-stakeholder reference
Structured as a shared resource for DPHE, ITN-BUET, UNICEF and the Bangladesh WASH Cluster, not one agency's internal tool.
Naya in action
A real exchange with the live assistant.
This is an actual conversation with Naya on the public site, captured for this case study — abridged for length, unedited in substance. The costs it cites match the technology database directly.
Question asked
“I need a budget estimate for providing safe water supply to a flood-affected char area with about 300 households. What technologies would you recommend and what would it cost?”
Naya — WASH Assistant
For a char that experiences regular flooding, the Dual-Platform Tubewell is the most cost-effective and resilient primary solution — the pump head sits on a platform that stays above flood level, and it needs no electricity. A Ring-well can be added where arsenic is a concern; a Tara tube-well can serve as a deep backup source.
Scenario A — all Dual-Platform Tubewells
Dual-Platform Tubewell9 units × ৳52,123৳469,107
Annual O&M (≈5%)৳23,455
First-year cost৳492,562
What the live system shows
A working proof that grounded AI can hold real engineering judgment.
Every cost figure the assistant returns traces back to the technology database, not a guess generated on the spot.
Recommendations account for flood exposure, arsenic risk and seasonal water-table depth — not just a generic technology list.
Multi-technology, multi-scenario cost breakdowns compute correctly, from unit cost through O&M to first-year total.
In ordinary public use today, not a concept or a slide deck.
Where this could go
A foundation to build WASH technology planning on, nationally.
The system already runs the full loop — catalogue, cost, location and grounded AI guidance. The opportunities below build on that foundation.
Deeper agency integration
Connecting the assistant and catalogue more directly into DPHE and partner agencies' own planning workflows.
Procurement-linked budgeting
Connecting the assistant's cost estimates to live procurement or tender data would keep every figure current automatically.
A larger technology catalogue
Additional WASH technologies and regional cost variations could be added as the sector's own reference evolves.
Wider disaster-response reuse
The same grounded-assistant pattern could extend to other national technology-reference needs beyond WASH.
Evidence and framing
Grounded in the live public system.
This case study describes the public system at wash.dphe.online as directly observed, including a real exchange with its assistant.
The hero illustration is a representative concept; the budget figures shown match a real exchange with the live assistant. No confidential data or internal system screens are reproduced here.
Grounded AI, built and running