Building JalSetu: An Event-Driven, Serverless Water Equity & Delivery Platform on AWS using Bedrock & Amazon Location
Discover how JalSetu leverages Amazon Bedrock (Nova Lite), Amazon Location Service, Step Functions, and DynamoDB Streams to bring transparency, algorithmic route optimization, and cryptographic proof-of-delivery to urban water crisis management.
JalSetu (जल–सेतु): Architecting an Event-Driven, Serverless Water Equity & Delivery Platform on AWS
"Understand water needs. Prioritize transparently. Plan the road ahead."
Water scarcity in rapidly expanding global metropolitan areas is not merely an infrastructure deficit it is an information, logistics, and governance challenge. In dense megacities like Delhi, millions of residents depend daily on emergency municipal water tankers. However, the existing tanker distribution ecosystem faces three structural hurdles:
- Opaque Prioritization: Emergency triage is often driven by manual grievance redressal without reproducible mathematical equity.
- Constrained Urban Fleet Routing: Heavy multi-thousand-litre tankers navigate dense street networks without automated, capacity-aware truck routing.
- The Proof-of-Delivery Dilemma: Proving that water was legitimately delivered to the intended vulnerable community without diversion, ghost deliveries, or false claims—has historically lacked cryptographic verification.
To solve this, we engineered JalSetu (जल-सेतु): an event-driven, serverless platform built on AWS that bridges vulnerable residents, municipal dispatchers, and tanker fleets through algorithmic transparency, Amazon Bedrock AI triage, Amazon Location Service road routing, and Step Functions proof-of-delivery sagas.
- Live Hosted Pilot: main.deemkw95tpvmo.amplifyapp.com
- Open Source Repository: github.com/mysterio-Apoorva/JalSetu
🧭 The Production Reality Boundary
Unlike typical hackathon prototypes that claim magical capabilities, JalSetu is built on engineering honesty. What is live, what uses real AWS services, and what is safely simulated are strictly delimited:
Feature | Source or Mechanism | What the Result Means |
|---|---|---|
Authentication & RBAC | Amazon Cognito User Pools | Real accounts; public signup grants Citizen access; administrative assignment for Driver , Dispatcher , Auditor . |
Search & Route Geometry | Amazon Location Service (HERE) | Real provider road geometry; server resolves origin; truck travel profile enforced; no straight-line fallbacks. |
Weather Integration | Open-Meteo Historical Model API | 30 complete days of precipitation and temperature data; excludes incomplete current-day records. |
Population Density | WorldPop Global 2 (Year 2025) | High-resolution modeled population within a 0.02° bounding grid cell; attributed under CC BY 4.0. |
Triage & Operations Copilot | Amazon Bedrock ( nova-lite-v1:0 ) | Language model reasoning on incident complaints and operational fleet snapshots; isolated from resident scoring. |
Resident Priority Score | Server-Side Formula ( pilot-priority-v2 ) | Deterministic, auditable formula ensuring zero algorithmic hallucination in civic resource allocation. |
Trip Confirmation OTP | Amazon SES | Real 6-digit cryptographic verification code sent to verified resident email; rate-limited and salt-hashed. |
Physical Delivery | Boundary of the Pilot | Fleet movements and water volumes require municipal onboarding; no invented delivery numbers. |
🏗️ System Architecture & Event-Driven Topology
JalSetu is designed from the ground up as a cloud-native, serverless, and event-driven architecture (EDA). The entire stack scales to zero when idle, enforces sub-second latency for resident requests, and isolates operational workflows into resilient asynchronous sagas.
A SNEAK PEAK:
🔬 Rigorous Automated Verification
Ahead of submission, JalSetu was subjected to end-to-end regression validation:
Verification Layer | Test Scope | Result |
|---|---|---|
Backend Regressions | Input validation, JWT role enforcement, OTP hash security, atomic DynamoDB settlement. | 67 / 67 Passed |
Production Browser Tests | Playwright Chromium end-to-end user journeys, map rendering, and error boundaries. | 9 / 9 Passed |
Live AWS Service Probes | Verified Amazon Location Vasant Kunj route (17.559 km, 407 geometry points) and Bedrock Nova Lite copilot. | Verified Live |
IaC Template Validation | CloudFormation / SAM template parity, IAM least-privilege scoping, and strict CORS. | Verified Clean |
🌟 Key Architecture Lessons Learned
- Hybrid AI Engineering (Deterministic Math + Generative Intelligence): The most impactful architecture decision was refusing to let an LLM calculate numerical civic priority scores. Combining deterministic formulas for resource allocation with Amazon Bedrock for qualitative triage gave us ethical, auditable, and hallucination-free governance.
- Step Functions Task Tokens for the Physical World: The
waitForTaskTokenpattern bridges cloud automation with the real world. Instead of running expensive server workers to monitor truck arrivals, Step Functions sleeps until awakened by an authenticated driver payload.
- Atomic Writes Guarantee Integrity: In civic logistics, duplicate delivery claims drain public resources. Leveraging DynamoDB's
TransactWriteItemsensured that tanker inventories, citizen quotas, and audit ledgers update atomically or fail together.
🔗 Try JalSetu
- 🚀 Explore the Live App: https://main.deemkw95tpvmo.amplifyapp.com
- 💻 Inspect the Source Code: https://github.com/mysterio-Apoorva/JalSetu
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article