Nakshatra Shah

I live in

contrasts.

Chartered Accountant. Self-taught builder. Stage performer since age seven. Currently at JP Morgan, Mumbai. Figuring out what's next, loudly.

scroll

Now · live

"Head in a book at home.

A microphone in my hand on stage."

Five contrast pairs scroll past as you read down. Each one is proven, not claimed.

1 / 5

Now

Working at

JP Morgan

Product Control, Treasury & CIO desk. Mumbai.

Reading

The Return of the King

Tolkien, third time round, still cries.

Building

Something I haven't announced.

Check back in a few weeks.

Watching

Cricket. Always.

Mid-digital-detox on everything else.

Thinking about

[[ ONE BIG QUESTION ]]

[[ EDIT ME — replace with what you're sitting with this week ]]

01 / 05

Auditor

Three years at EY. Statutory and SOX audits across NBFCs, asset management, capital markets, AIFs.

A ₹300 Cr public NCD issue audited line by line. Borrowings and debt securities — the second-largest line on an NBFC's balance sheet — re-engineered into a working paper that ran 70-80% faster than what I inherited.

Picked up a tool nobody else on the team wanted to learn (DataSnipper), got fluent, then trained the rest of them.

Failed CA Final on my first attempt. Cleared it on my third. The honest version of resilience is "kept showing up."

Builder

Two weeks. Mostly nights. Many bugs. Many hours waiting for AI usage limits to reset.

NettWorkk Predictor runs on a Cloudflare Worker I deployed myself, with a D1 database, real-time odds powered by Apps Script triggers, mobile-optimised. 50 users on a system designed for 200.

Last year I ran the same predictor league on Excel and Google Forms, copy-pasting entries by hand. This year I refused to.

I won't pretend I wrote every line. I'll tell you exactly which parts I built, which parts I directed, and how I think about it.

[[ CONNECTOR — EDIT ME. Suggestion: "Audit taught me how to read a system. Building taught me how to make one." ]]

02 / 05

Numbers

99 in Accountancy. 98 in Maths. 96.80% in 10th. 92.15% in 12th.

CA Final cleared, both groups. CAT 97.15 percentile, taken while working full-time at EY. CGPA 8.76 at Narsee Monjee.

Numbers have always been a language I speak fluently — not because they're easy, but because they don't lie.

Stage

I was seven when I played Krishna in Krishnaparva — a stage drama with over a hundred artists, grand for its time. The lead role.

Lead in Aav Re Varsaad. Protagonist in the Hindi film Main Gaurakshak — small-budget, made by a Jain organisation to spread awareness on cow protection, never theatrically released. The work was real.

Winner of intra-school and inter-school elocution competitions. Public speaker, debater, anchor.

A microphone is a different kind of math. The variables shift, the answer is in the room.

«PHOTO 2 — STAGE MOMENT»
Krishnaparva / speaking / any old photo

[[ CONNECTOR — EDIT ME. Suggestion: "Numbers were the discipline. Stage was where I learned to say something that mattered." ]]

03 / 05

Quiet

[[ THE QUIET YOU — WRITE YOURSELF ]]

What you read, how you think alone. What books have you reread? What kind of writer pulls you in? When you're alone for a long evening, what are you actually doing?

Loud

[[ THE LOUD YOU — WRITE YOURSELF ]]

On stage, behind a microphone, in a room. What changes in your body when you walk on stage? What kind of speaker do you become? What did twenty years of speaking teach you that you couldn't have learned any other way?

[[ CONNECTOR — EDIT ME ]]

04 / 05

Plan

Five years for the CA. Multiple attempts. The kind of preparation that teaches you what consistency actually looks like — not motivation, just showing up to the same chair every day for years.

Then 97.15 percentile on the CAT, on first try, while working at EY.

Three years of audit deadlines that don't move because you're tired.

Improvise

NettWorkk started as a half-formed idea. I didn't know JavaScript. I didn't know Cloudflare. I didn't know what a worker was when I started.

I learned by building. Used AI tools as a thinking partner. Hit usage limits, waited hours, tried again. Burned weekends on bugs that turned out to be typos.

Two weeks later it was live.

[[ CONNECTOR — EDIT ME. Suggestion: "Plans got me here. Improvisation is what I do once I trust myself." ]]

05 / 05

Earned

I grew up in a house where 94 out of 100 meant try harder. Where the path was Commerce, then CA, then a respectable firm, then security.

I followed it. I'm grateful for what it taught me — rigor, patience, the muscle of doing hard things slowly and not flinching when they don't work the first time.

Chosen

I'm not staying on that path.

I want to build, or join people building. I want my work to be useful to a small team rather than reviewed by a large one. I want to be in rooms where the question is what should we make, not did this comply.

The thing I earned gave me the tools. What I'm choosing is what I do with them.

[[ CONNECTOR — EDIT ME. The landing line of the whole thesis. Suggestion: "Both sides of this are me. The contrast isn't a problem to resolve. It's the point." ]]

Built

Three things I'm proud of, told honestly.

Case · 01

Built

Built in 2 weeks. Mobile-first. 50 users on a 200-capacity system.

NettWorkk Predictor

A live cricket prediction league running on Cloudflare Workers.

NettWorkk started as a side project with my brother — a community-driven marketing company we were trying to build. We never quite found the spark. The bottlenecks were real. We didn't take off.

But one thing kept going: a small IPL predictor league we ran for friends and a small community.

Last year I ran it on Excel and Google Forms. Every match, I'd download form responses, copy-paste them into a sheet, recalculate the leaderboard by hand, post it back. Hours of work. Errors I had to chase.

This year I refused to do that again. Not because I was a builder — because I was tired of being the manual labour my own project required.

nettworkkpredictor.pages.dev

Status

Live

on Cloudflare

Users

50 / 200

capacity designed for

Built in

2 weeks

mostly nights

Cloudflare Worker · D1 · Apps Script · mobile-first

An IPL prediction league built in JavaScript, hand-deployed.

Open the live site

Live numbers update when an endpoint is exposed. Static for now.

The build

I'd never written production code. I started anyway.

Two weeks, mostly nights. A Cloudflare Worker in JavaScript. A D1 database for users and matches. Apps Script triggers on the Google Form so that when someone submits a prediction, the data flows to the Worker, which updates the live odds and leaderboard.

I worked with Claude as my pair programmer. Hit usage limits. Waited hours. Used other AI models in the gaps. Burned weekends on bugs that turned out to be one-character typos.

I'll be straight: I don't write code in the way an engineer does. I have ideas, I direct AI tools, I integrate, I debug, I ship. The distinction matters and I won't blur it.

What I learned

Building is mostly waiting and not giving up. The romantic version where you sit down and code for six hours and ship — that's not how it works for me, and I suspect not for most people. The unromantic version is two weeks of small, frustrating, eventually-working steps.

That's enough for me. I'd do it again.

Case · 02

Built

Three years at EY, in plain language.

Audit, Re-engineered

The work paper that became the team's template.

The problem

When I joined the engagement, the work paper for borrowings and debt securities — the second-largest line on an NBFC's balance sheet — was a mess. Unstructured client data on one side, manual re-keying on the other, fourteen days to audit it, and most of those days went to copy-paste. The numbers were right. The process was wasteful.

The fix

I rebuilt the work paper from the ground up. Standardised input formats, formula-linked tabs, named ranges, dynamic reconciliations across statements of accounts from multiple lenders, automated EIR and amortisation calculations as per Ind AS. No macros, no VBA wizardry — just disciplined Excel applied with intent.

The result: 70-80% less manual work. Days back for actual analysis. A template that became how the team handled the area afterwards.

While I was at it, I taught myself DataSnipper — a verification tool the firm had introduced that nobody on my team had explored. Within weeks I was training the rest of the team. We cut document verification time roughly in half.

The other stuff

There was more. In year one, I picked up an entire engagement that had been abandoned by someone two levels senior to me, and got recognised for it. I worked on a ₹300 Cr public NCD issue. SOX audit for a global financial services firm. The usual statutory work across asset management, capital markets, AIFs.

I was a team player. I had good bonds with everyone. I was good at audit because I was attentive, because I built tools that made my work better, and because I noticed when the existing system wasn't serving us.

Those instincts didn't come from audit. Audit just happened to be where I sharpened them.

The honest part

I didn't love it. I'll say that here because pretending otherwise is exhausting and you'd see through it.

What I loved was the part where I made the work better. That part I'd do anywhere.

Case · 03

Built

Six years of side projects in language.

Words

Writing and translation, the throughline I almost forgot to mention.

Clienter.AI — Blog Writer

Six months writing for an AI accounting startup. Authored long-form posts on AI trends, automation in accounting, and user education for a non-technical audience. Read the work →

Mangrol Multimedia — Translation

Four years of freelance Gujarati-to-English translation work. A UPSC exam guide book. Two autobiographies — one I led, one I assisted on. Form digitisation projects for an MNC bank (35+ forms converted into editable PDFs).

Translation taught me what writing is when meaning matters and the wrong word costs something real.

This site

The words on these pages — most of them — are mine. That counts as a portfolio entry too.

Next

where this goes

[[ THE BRAVE STATEMENT — WRITE YOURSELF ]]

Suggested structure (delete once you write):

Paragraph 1. What you actually want. MBA at a top-10 program. Joining a startup early. Founding something with the right co-founder. All three on the table. Be specific about which schools, which kind of startup, which problem space pulls you.

Paragraph 2. Why. What kind of problems pull you. What kind of teams you want to be on. The kind of work that would make you say yes on the first call.

Paragraph 3. An invitation. What would make you reply to a cold email. Who you're hoping reads this and writes back.

Length target: 250–350 words. Conversational, not corporate. Write it like you're emailing one person, not posting to LinkedIn.

Always

~600 words

Photo

«PHOTO 1 — HERO PORTRAIT»
head & shoulders, calm gaze

I live in contrasts.

I'm often quiet at home, head in a book, the kind of evening that ends with three more chapters than I planned. I'm also expressive behind a microphone, on stage, at the front of a room. I've spent twenty years moving between those two modes and I've stopped treating it like a contradiction.

I grew up in a house where 94 out of 100 meant try harder. The expectation wasn't cruel — it was clarifying. Aim higher, prepare better, don't settle for almost. I chose Chartered Accountancy partly because of that ethic and partly in spite of it. CA is a profession known for failure rates, and I picked it anyway, because the only way I knew how to test myself was against something hard.

It tested me. I cleared CA CPT in 2019, Intermediate in 2021, my B.Com from Narsee Monjee in 2022 with an 8.76 CGPA, and CA Final in 2025 — the second part on my third attempt, after a first attempt that taught me what underprepared actually feels like. Three years at EY in between, on statutory and SOX audit teams across NBFCs, asset management, capital markets, and alternative funds. CAT 97.15 percentile in 2024. The numbers stack up neatly on a page; the years behind them did not.

Audit gave me something I'm grateful for and something I'm leaving behind. The thing I'm grateful for: the discipline of looking at any system and figuring out what's actually true beneath what it claims. The thing I'm leaving behind: the part where the answer is "did this comply" instead of "what should we make."

The other half of me has been on stage since I was seven. Krishna in Krishnaparva — a play with over a hundred artists, the lead role, the kind of memory you carry. Aav Re Varsaad after that. The Hindi film Main Gaurakshak. Years of elocution wins, debate, anchoring. A microphone, I learned early, is a different kind of math — the variables shift mid-sentence, the answer is in the room, and being right matters less than being heard.

In between, I've worn other hats. Team Leader at Viral Fission running 35 campus ambassadors. Blog Writer at Clienter.AI on AI and accounting. Joint Treasurer at a Leo Club. Translator on freelance projects, including two autobiographies from Gujarati to English. Marketing co-founder of NettWorkk with my brother — which didn't take off, but kept its predictor league alive long enough that I rebuilt it this year on Cloudflare Workers, two weeks of nights, my first real product.

I'm at JP Morgan now, in Product Control on the Treasury and CIO desk. Mumbai. Mid-2026. The next chapter is the one I'm still writing — maybe an MBA at a top-10 program, maybe a startup as employee number ten, maybe my own thing with the right co-founder. Honestly all three are on the table, and the truth is I'm optimising for the work that makes me feel most alive when I'm in it.

I'm a curious person. I solve problems with clarity, logic, and creativity, and I think the rarest skill in any room is empathy combined with rigor. I'm always learning. I'm always building. I'm always looking for the next thing that pushes me deeper.

If you've read this far, we should probably talk.

[[ This essay is a v1 draft. Rewrite it in your own voice when you have time. ]]

Reach out

If you're building something that matters,
I want to hear about it.

Last updated 1 May 2026 · Built in Astro · Hosted on Cloudflare

Try the OS version → coming soon