My Story

Where I grew up, the internet came and went like a guest. The problems stayed.

01

The Village

I was born in a village in West Bengal. My family lives there still.

Evenings smelled of wood smoke. The lamp flickered when the power cut came. Phone signal lived on one rooftop, and everyone knew which one.

Nobody called it a hard life. It was just life.

02

The Questions

I watched my neighbours. A farmer held a yellow leaf up to the sun and guessed. Was it a disease or just a dry week? By the time he knew, half the field was gone.

I watched a sick man travel hours to learn that a small check would have been enough.

Smart tech was real. But it was made for big cities, with fast internet and tall buildings.

One question followed me everywhere: Why can't it come here too?

03

The Wall

I started with a wearable health monitor. It had to catch danger early.

But the small CPU was slow. Each result took over 800 ms. That is too late when a patient needs an alert.

Then I faced the harder truth. A cloud system needs the internet. My village does not have it every day. The idea was dead before it began.

04

The Fix

So I stopped sending the model away. I brought it home.

I put it on a Jetson Orin Nano, a small board with a strong GPU. I shrank the model with FP16 and INT8 quantization. I used CUDA to make it fast.

The result: under 100 ms. More than 4x faster than the Raspberry Pi. No cloud needed.

The first time it ran in a dark room with no signal, I just stared at the screen. It worked. Right there. On its own.
  • <100 msOn-device inference
  • 4×+Faster than Raspberry Pi
  • OfflineNo cloud dependency
05

The Field

Then I turned to the farmer with the yellow leaf.

I trained a model on 1,500+ field images. It reached 96%+ accuracy in finding crop disease. I added soil checks, a fertiliser calculator, weather data, and a market for farm products. I made the dashboard simple enough for someone who has never used a dashboard.

Now the leaf gets an answer before the field is lost.

  • 1,500+Field images used for training
  • 96%+Crop disease accuracy
  • 3+Farms using the system
06

The Village Thread

Farming. Health. Banking. Education.

I built a learning platform for 200+ rural students. I built a banking system with fraud checks for rural users.

Different fields, one reason: good tech for people far from the city.

  • 200+Students on the learning platform
  • MLFraud detection in rural banking
  • 4 fieldsFarming, health, banking, education
07

What's Next

My family is still in the village. So I always know who I am building for.

More edge models. More farms. More real users.

Got a village problem?

Bring it to me.

Let's talk