Skip to the main content
PROTOTYPEA student prototype. Not an official government service.

What is real and what is pretend

This is a student prototype. Here is exactly where the line is.

Real in this prototypePretend
The whole journey, start to finishAadhaar number, the code on the phone, the PPO lookup
Taking the photo and checking its qualityMatching your face against the government database
The plain-language explanation when something needs fixing (live OpenAI call)The pension office system that gives the result
The check on your photo before it is sent (live OpenAI call)The message to your phone (written to a visible outbox instead)
Spoken guidance in every language here, including the ones no phone has a voice forBank and post-office links
The status state machine and its audit log

See the messages we would have sent

In real life this goes to

What do you need to do?In real life this goes toThe real form is
Old-age pensionnsap.nic.inNSAP application form
Widow pensionnsap.nic.inNSAP application form
Disability pensionnsap.nic.inNSAP application form
Pension from your PFunifiedportal-mem.epfindia.gov.inForm 10D
Government service pensionbhavishya.nic.inForm 6-A
Put money aside now for a pension at 60npscra.nsdl.co.inAPY subscriber registration form
Free grain every monthnsap.nic.inNSAP application form
Take over a pension after a deathpensionersportal.gov.inForm 14
Money after a death in the familynsap.nic.inNSAP application form
Prove you are alivejeevanpramaan.gov.in
Move your pension to another bankcpao.nic.inPension transfer request
Extra pension after 80pensionersportal.gov.in
Get back the part of your pension you soldcpao.nic.inRestoration request
My pension has not comepgportal.gov.in / CPENGRAMS

How this would work at real scale

Where it plugs in. This is a front end. In production it would call the existing Jeevan Pramaan and UIDAI face-match services through an authorised pension-agency integration — the government stays the system of record. Nothing here duplicates a government database.

What we would never store. Aadhaar numbers, fingerprint and face data, and photos are all transient. The photo goes to the matching service and is discarded; only the resulting reference number and status persist.

The AI is advisory only. It never decides an outcome. It only explains an outcome the government system already returned, and warns about a photo before it is sent. A wrong AI guess costs one retake, never a pension.

Falling back to humans. Every dead end routes to a phone number and a list of nearby help centres. Digital-first must not mean digital-only.

Scale. Roughly one crore submissions a year, heavily concentrated in November and December. That is a queue problem, not a compute problem — the async job model used here is the shape of the answer. Staggering reminders by PPO series would flatten the peak.

Accessibility as infrastructure. Eleven languages here, and the twelfth is a data file: one row in the registry and a dictionary that may start out half-written, because every string falls back rather than breaking. Twenty-two scheduled languages is a content problem, not a rebuild.

For the technically curious

Next.js 15 App Router, TypeScript, Tailwind v4. The mock pension office is a state machine with an audit log, behind real HTTP route handlers, backed by an in-memory Map — in production a Postgres table and a job queue. The OpenAI model gpt-4o-mini is called only from the server, only for the two jobs above, and every call has a hardcoded fallback so the app works with the key removed.

Pension Saral is not affiliated with, endorsed by, or connected to the Government of India, MeitY, UIDAI or Jeevan Pramaan. No government emblem, logo or flag is used anywhere in this prototype.