Nishit Bhatt

Hi, I'm Nishit. A product manager who ships the thing.

much better ✓

I live in Bengaluru. I helped build myBillBook, billing software for India's small businesses, from its early days. Lately I build AI products too, like heyAnaya, which answers clinic patients on WhatsApp and books their appointments.

I like turning messy real life into software that works, and then checking that it actually did. I write about that here.

the hard part

Explore my work ↓  ·  Talk to me ↗

Bengaluru · building heyAnaya

WorkOpen any to see it

the detail is in the decisions

  • Illustrative conversation · fictional patient · no live bookingExplore heyAnaya ↗
    The problem
    A patient messages after closing time. They still need an answer and a clear next step.
    My part
    Product rules and hands-on building across WhatsApp, booking, and staff handoff.
    The decision
    Move routine requests forward. Leave uncertain clinic details for staff to confirm.

  • Sharma Traders

    Tax invoice · 24 Oct

    Basmati 25kg ×46,400
    Toor dal 10kg ×34,200
    GST 5%530
    Total₹11,130
    SupplierSharma Traders
    Items2 items
    TaxGST 5%
    Total₹11,130
    Illustrative scan · sample bill and fieldsExplore myBillBook ↗
    The problem
    A shop owner has a pile of supplier bills to enter, item by item.
    My part
    I run the Scan & Record product and its pricing at myBillBook.
    The decision
    Turn a photographed bill into book entries. Work through item, tax, and total rules before treating extraction as done.
  • The machogs website
    Signed, notarized, and never closes anything without asking.Get machogs ↗
  • Shelf: Save now. Understand later.
    Save now. Understand later.Open Shelf ↗
  • ~ $
    Fetching bnishit/skills…
    ✓ Ready in every agent on this machine
    Illustrative install · shared instructions for the agents I use.View on GitHub ↗
  • Tide: Find your flow.
    Forgiving workout logs. Running stays optional.Try Tide ↗

Try it · the front desk3 small decisions

the reply is the product

What would you send?

A patient messages after the clinic closes. Pick a reply and see what it changes. No timer. You can try every answer.

Fictional clinic and messages. Nothing is sent or booked.

NowUpdated 27 Oct 2026

the kettle's on

Building
heyAnaya, helping clinics in the UAE answer patients on WhatsApp and follow through on appointments.
Reading
Saved to Shelf this week: Memory in Voice Agents Is a Harder Problem Than You Think, and The Token Uplift, on pricing AI work.
Thinking about
What a PM is for when building is cheap. Short answer so far: deciding what not to build.
Away from the screen
Swimming. I built Tide to keep the logs kind.
On my desk
Claude Code and Codex as a daily build team, Obsidian for notes, and Shelf for everything I mean to read.

Writing

    Shelf · lately read and watched

    pull one out ↓

    How I work

    still relearning these, weekly

    1. Edge cases are the product. A part-payment with a discount and tax deducted at source isn't a corner case for a wholesaler. It's Tuesday.
    2. Ask what the user will feel. Dashboards show what people did, not the patient who felt ignored and quietly left.
    3. Evidence over claims. A passing test doesn't mean the message reached the patient. I check what the user sees.
    4. Decide what not to build. When building is cheap, the scarce skill is saying no to nine ideas and being right.
    5. Build the smallest real thing. A prototype on real data often beats a spec. AI makes that cheap for a PM now.