Address intelligence for last-mile delivery

Right address.
Right office.
First time.

PostAI turns incomplete or inconsistent addresses into a clear, explainable delivery-office recommendation—before a parcel enters the wrong route.

30demo offices
4ranking signals
100%explainable
Route check preview Address analysis
Ready
Destination 24 Gandhi Road, Anna Nagar, Chennai 600040
01
Address components extractedDone
02
PIN and locality verifiedDone
03
Nearby offices rankedDone
Best delivery office
Anna Nagar H.O.
600040 · 0.1 km away
Highconfidence
The routing gap

Bad address data becomes a delivery problem.

A single wrong digit or ambiguous locality can send a parcel into manual review. PostAI surfaces that uncertainty while there is still time to act.

01

Catch inconsistencies early

Compare the stated locality with the PIN code’s expected service area before routing.

02

Reduce manual lookup

Turn unstructured address text into normalized components that are easier to review.

03

Make decisions auditable

Show the exact PIN, distance, locality, and office-status signals behind every result.

Incorrect PIN Missing locality Ambiguous area name PIN–area mismatch Incomplete address
How it works

One address. Five visible decisions.

Watch a raw address move through PostAI’s complete route check—from capture and validation to a ranked, explainable handoff.

  1. 01
    Input

    Capture the address

    Accept typed text or a clear image of a printed address label.

  2. 02
    Structure

    Parse components

    Separate the house, street, locality, city, state, and PIN.

  3. 03
    Verify

    Check PIN and locality

    Flag missing details or inconsistencies that require review.

  4. 04
    Locate

    Measure proximity

    Geocode the destination and score nearby delivery offices.

  5. 05
    Recommend

    Rank and explain

    Present the strongest candidate, confidence, and supporting evidence.

Inside the model

Clear inputs.
Visible reasoning.

PostAI combines deterministic parsing, geospatial distance, and a transparent weighted score—without hiding the route decision.

System architecture

Browser-based prototype
InputAddress text or label
↓
UnderstandParser + PIN validator
↓
LocateNominatim + local fallback
↓
CompareOffice dataset + distance engine
↓
OutputRanked, explainable recommendation
Scoring model

Four signals, one recommendation.

40%
PIN match
Exact, sub-zone, or no match
30%
Distance
Haversine distance in kilometres
20%
Locality match
High, partial, or no match
10%
Office status
Active or inactive
Formula 0.40 × PIN + 0.30 × distance + 0.20 × locality + 0.10 × status
Open geospatial stack

Built from practical, inspectable parts.

The prototype uses public mapping tools and a local coordinate fallback for common Chennai localities.

Geocoding

Nominatim

Converts a postal address into latitude and longitude.

Mapping

Leaflet + OSM

Plots the destination and top candidate offices on an interactive map.

Fallback

Local area lookup

Provides approximate coordinates when live geocoding is unavailable.

Sample analytics

Operational signals,
made visible.

A complete view of prototype outcomes, confidence levels, mismatch patterns, and routing distance. Figures are demonstration data.

Address → validate → locate → recommend
Demo data

These figures illustrate the product experience and do not represent live India Post statistics.

01Sample total
128

Addresses analyzed

02Sample flag rate
17

PIN mismatches flagged

03Sample success rate
109

Successful matches

04Sample average
91.4%

Average confidence

Confidence distribution

How the 128 demo results are distributed
All results

68 results have very high confidence, 29 high, 18 medium, and 13 low.

Monthly analysis volume

Prototype sessions from February to August
7 months

Monthly demo volume rises from 12 in February to 27 in August, totalling 128.

PIN consistency by area

Matched and mismatched addresses in top localities
Top 6

Matched addresses outnumber mismatches in all six displayed localities; Anna Nagar has the highest volume.

Distance to recommended office

Average monthly route distance in kilometres
Lower is better

Average distance falls from 4.1 kilometres in February to 2.4 kilometres in August.

About the prototype

Designed for smarter postal routing.

PostAI explores how address quality, geospatial context, and transparent scoring can support faster delivery-office identification.

Smart India Hackathon

Problem statement SIH1758

Software prototype
Challenge
AI-powered delivery post office identification
Organization
Ministry of Communication
Department
Department of Posts
Theme
Transportation and logistics
Problem

Incorrect, incomplete, or ambiguous addresses create manual lookup, misrouting, and delivery delays. The prototype tests whether address parsing and geospatial matching can identify the most likely delivery office and flag uncertainty for human review.

Design principles

Useful before impressive.

The prototype prioritizes explainability, low-friction review, and tools that can be inspected or replaced.

01

Structured, not mysterious

Pattern-based extraction keeps every detected address component visible.

02

Geospatial by default

Office proximity is measured rather than inferred from PIN data alone.

03

Human review stays possible

Confidence and mismatch signals make uncertain decisions easy to spot.

Prototype stack

Small, open, and browser-based.

HTML + CSS
Responsive interface
Vanilla JavaScript
Parsing and orchestration
Leaflet
Interactive map
OpenStreetMap
Map tiles and geocoding
Chart.js
Prototype analytics
Tesseract.js
In-browser OCR
Try the prototype

Put an address through the route check.

Use a Chennai address or load one of the built-in scenarios.

Analysis workspace

Resolve an address.

Enter text or upload an address label. PostAI will parse, validate, locate, and rank the most likely delivery offices.

Address analysis results

Waiting for an address

Your route check will appear here.

Start with a prepared scenario or enter your own Chennai address.