SayFlow
Hold a key. Speak. Let go.
Your words land in any Mac app.

I listen, build, and own what ships.
From enterprise AI to the apps in your hand.
Sometimes it’s a bank processing millions of documents. Sometimes it’s my grandmother checking her phone. The scale changes. The responsibility doesn’t.
01 enterprise product lineInvoices. Forms. Real customer work.
Several models read each document. A fixed rule decides what can be automated and what needs a person.
Several models read the same document independently. This illustration does not run the production system.
Four different worlds. One reason to build.
My grandmother is 75. She wanted one simple answer: when is Rahu Kaal today?
Three versions. Each simpler than the last.
Now she checks it on her own.
Large type, full Hindi, and voice callouts. Designed around how she actually holds and uses her phone.

For anyone who’s stood in a gym pretending to stretch while secretly googling a machine.
Point your camera. Understand the equipment.
Walk in with a little more confidence.

Last night’s dishes shouldn’t be the first thing you face. One photo before bed. A vision model checks that the sink is actually empty. Your streak grows a star at a time.
Take one photo of your empty sink before bed.

Running on my phone. Kept under wraps on purpose.
If you’re curious about what I’m building, let’s have a conversation.

Working betas, born from small frustrations. A few things currently on my desk.
Hold a key. Speak. Let go.
Your words land in any Mac app.

Good things, going free.
Find what’s nearby, in one place.
A resume that starts with a conversation, not a blank page.
Calorie tracking that starts with your plate. Just take a photo.
I’m Kanishk. An AI product manager who likes being in the room with customers—and staying until the thing actually works.

The best part of building?
Watching someone use the thing
without needing you to explain it.
At AWS, that meant listening to enterprise accounts already on their way out. At Itemize, it meant catching a healthcare customer’s failing claim forms. At home, it meant watching how my grandmother held a phone.
I start there. Then I work backwards to the product, the technology, and the thing we can actually ship.
Three versions of RahuKaal. Each simpler, until she could use it alone at 75. I get blessings in return.
Selling is listening until the value is obvious, then saying it plainly. That’s how I approached at-risk accounts at AWS.
Taking AI from idea to production. Measuring it honestly. Selling what I ship. Breadth makes the whole thing possible.
Cloud. Insurance. Education. AI.
Find the person behind the number.
Then build for them.
Document AI for healthcare, banking, and finance. I own the prompts, evals, architecture decisions, and go/no-go on each release.
Worked with legal, marketing, and sales to turn an AI adoption mandate into a funded architecture and a production sales chatbot.
Managed a $50M+ Fortune 500 portfolio across 25 enterprise accounts. Retained at-risk accounts, grew Enterprise Support, and led cost-optimization sessions.
Launched a tablet learning platform through a $1M Lenovo partnership. Grew the user base from 3.5M to 6.5M in eight months.
Took a digital education platform from zero to MVP in five months, with engagement up 35% through data-driven iteration.
Ten credentials across AI, cloud, and delivery, so I can speak the buyer’s language and the builder’s in the same meeting.
AWS Solutions Architect Associate
AWS SysOps Administrator Associate
AWS Cloud Practitioner
AWS Cloud Economics Essentials
Google Cloud Digital Leader
PMI Generative AI for Project Managers
Project Management Professional
Certified Associate in Project Management
ITIL 4 Foundation
Certified ScrumMasterSmall releases, production milestones, and the occasional stubborn bug.
I designed the system, wrote every word, and directed AI to build it, plus a chat that only knows my real work. Every claim on this site is a thing that shipped.
The daily window Hindu texts say to avoid, set by the sun, plus the sandhya prayer windows, computed for the city you ask about, with elder-first onboarding, full Hindi, and voice callouts. Conceived, designed, and shipped solo, twice: for iPhone and for Android. Its page here computes your own Rahu Kaal window, live.
Wake up to last night's dishes? Photograph your empty sink each night; a vision model, not the honor system, decides whether it counts. This pass, on device: Lock Screen widget, Live Activity, a rescue override for hard nights, and a flood easter egg for people who find it.
For anyone nervous to enter a gym: aim at any machine and it tells you what it is and coaches the movement in two or three lines, no stranger to ask. This release: a full design pass, a new app icon, and a stubborn microphone bug hunted down and fixed for good.
Scaled from pilot to the production path for enterprise document volume. Human review now touches the exceptions, not the pipeline.
Enterprise banks, payment processors, and healthcare customers needed financial documents classified at production accuracy. One product manager, one week: an LLM ensemble behind a deterministic hard gate (several AI models read each document and vote; a fixed rule decides what is automated and what a person checks).
AI PM, Forward Deployed Engineer, and Deployment Strategist roles. Open to relocating anywhere in the US. Put me in front of customers. I’ll take it from there.