The Book
of Avidha

Once there was a girl from Guwahati, where the Brahmaputra is so wide that in the rain you can’t see the other side. She could never stay quiet. At St. Claret School they made her Head Girl, perhaps hoping it would keep her busy. It didn’t.

She went west to Odisha to study data science at IIM Sambalpur, and found out that numbers are most interesting when you go and meet the people behind them.

That girl is me. The rest of this book is about those people, and what I built for them.

Like everyone who watched Iron Man, I wanted a Jarvis. So I built one, minus the hologram. It lives entirely on my laptop. I can type to it, talk to it in English or Hindi, or wave at the webcam: pinch to spin its glowing core, open my palm to make it scan.

Getting it to run on Windows took longer than building it. Python was too new, the antivirus thought my tools were viruses, and my microphone was too quiet, so I taught it to listen to the room first.

Then I asked it a real question: how big a brain does it need? I tested a small model against one twice its size. The bigger one got one more answer right out of 24, and took half a second longer. I kept the bigger one, and wrote down why one answer isn’t proof of anything yet.

Everyone who gives money to CRY is promised a report: what happened to it, which children it reached, and when. There were many reports and many deadlines, and all of them lived in people’s heads.

When I joined as an operations intern, I gave them somewhere else to live. I built a shared tracker, and twenty-four little messengers that wake up on schedule, check what’s due, and knock on the right person’s door by email and on Teams.

Halfway through, the accounts I was building on stopped working. So I moved everything to one that belongs to CRY. The messengers still go out every morning, even though I’ve left.

In village after village, someone came and built a toilet, took a photograph and left. Then the soap ran out, and nobody came back to refill it.

During my month at Youth4Water Plus I kept asking who was supposed to come back. The answer was nobody. So I designed a business around the women who were already there.

In Swasthya Sakhi, women in self-help groups make soap and sanitary pads, sell them door to door as Hygiene Didis, clean schools on contract, and pack kits for the monsoon. They earn wages, a commission and a share of the profit, because they own it.

Bokaro is a city built around a steel plant that never sleeps. For a month I sat with the people who keep its computers talking to each other, and learned how a company that size remembers everything: SAP.

I wrote my first lines of ABAP, built reports and forms, and learned that nothing reaches the plant floor until it has passed through development, then quality, then production.

It was the least glamorous month of my year, and one of the most useful.

On paper, every village in India has electricity. But a child in rural Odisha might still do homework by kerosene lamp three nights a week. My friend Varun and I wondered who checks.

It turns out a satellite does. Every night NASA photographs the Earth’s lights from space and gives the pictures away for free. We taught a computer to compare each village with its own usual brightness, night after night, and to notice when one went dark while its neighbours stayed lit.

Then we tested it on the night Cyclone Fani came ashore in 2019, when Bhubaneswar fell to 11% of its normal light. It worked. We called it the Andhakaar Index.

At Golebazar market, an aloo-chop seller told us he boils potatoes until one o’clock in the morning. His wife peels beside him. Nobody pays her.

Every vendor on that street was doing the same work, alone, in the middle of the night. So Varun and I designed Taiyaari: one shared kitchen that preps for all of them, owned by the vendors themselves, one member, one vote. Each vendor still sells their own recipe at their own price. They just get to sleep.

When our first numbers said vendors would save 20%, we checked again and found it was closer to 10%. We told the judges 10%. We made the final round.

A weaver in western Odisha earns money every single day and has nothing on paper to prove it. So when she asks the bank for a loan, the bank says no.

But she has a phone, and she has WhatsApp. Hisaab lets her say what she sold today, in Odia, in a voice note. It turns that into a record of her income that she can show a lender.

We pitched it at IIM Sambalpur’s PRAYAS hackathon, for weavers, self-help group members and street vendors.

“Hello ma’am, this is Arjun from your bank’s security desk. Your card was just used for 48,000 rupees in Dubai.” Thousands of parents get a call like that, and some of them read out the OTP.

ScamShot makes that call first, on purpose. It rings your parent with a realistic, harmless scam, stops before anything is shared, and then explains, kindly, what gave it away.

The writing can be made big, it speaks Hindi as well as English, and it can talk slowly. At LaserHack 2026 we were picked for the Grand Jury.

I built an agent and gave it one rule: never say anything you can’t prove.

Give it the number of any company in Norway. It goes to the public registers and the company’s own website, and comes back with a profile. Every fact carries a receipt: the page it read, a fingerprint of that page, and the exact words it copied. Before it shows you anything it checks every receipt again, and anything that doesn’t match stays hidden.

When it can’t tell whether a website really belongs to the company, it says so. There’s no AI model inside it, so it has nothing to make up.

Before any of this, I wrote stories for children.

Echoes of the Starlit Phoenix is a collection of them: banyan trees that guard old secrets, gardens that grow upside down, and children like Deepak, Luma and Li Wei learning patience, kindness and courage from a jade phoenix and a silent monk.

I published it myself on Amazon, under the name MS Amaya. It’s the reason this portfolio is a book.