Rohith Avatapally, builder based in San Francisco, CA

San Francisco

Rohith Avatapally

Ex-Meta, ex-Amazon. Product first, whatever the stack.

I left Meta to slow down to human speed — we had our second kid, and I wanted to be present after a long stretch of FAANG and startup pace.

See the work →How I thinkWrite me
Amazon Music
10M+
podcast listeners reached
200K+
groups with GenAI starters
Olto
0→1
SaaS to paying customers
Mobile · TV · Console · Web
9
platforms shipped to

01 / work

Fifteen years, across backend, clients, and GenAI.

What I owned, the result, and the decision behind it.

Production GenAI in Facebook Groups — conversation starters that drove a confirmed +25K daily posts.

MetaEngineering ManagerDec 2024 – Nov 2025

I led 8 engineers in Facebook Groups. Starters shown to admins didn't land, and leadership had given the project about a 20% chance. We changed who did the posting: each group's own AI agent. It shipped to 200K groups and added 25K posts a day. We then moved that team to SEO discovery, and young-adult visits from Google went from 450K to 1M+.

+25K
daily post creations
200K+
groups at launch
450K→1M+
young adult visits from Google

Left eight years at Amazon to learn GenAI by shipping it — a 0→1 demo product to a paying customer in under six months.

OltoFounding Engineering LeaderJul 2023 – Dec 2024

I was the founding engineering leader and owned product and engineering from zero. Enterprise demos took days and went stale as soon as a prospect left the script, so we built a slides-style editor: capture the customer's live product, then let a sales engineer curate the demo with AI generated data for their industry. Setup got up to 70% faster, a paying customer landed in under six months, and I took SOC2 through certification.

<6 mo
to first paying customer
70%
faster demo setup
SOC2
certified end to end

See the product

Customer voice on the product, not a pitch deck. Click to play.Open on YouTube

In the press

Founding leader for Podcasts. Grew my teams from 7 to 20+ engineers and launched to 10M+ listeners.

Amazon MusicEngineering Manager2020 – 2023

I was a founding leader of Podcasts and ran three full-stack teams. We launched the browse, library, and playback verticals in about nine months, faster than Amazon Music had done it before. After that, we took podcasts to nine countries. Over the next couple of years I grew the teams I owned from 7 to 20+ engineers. Five of my ICs were promoted, including one who became a manager. The later launches I owned were ad-free podcasts for paid subscribers, then library engineering for Profiles.

10M+
active podcast listeners
7→20+
engineers on my teams

In the press

Moved Amazon Music's client logic to the backend, and cut a full new-client app from a team of 4–5 and 12 months to one engineer and 3.

Amazon MusicSr Engineer2015 – 2020

Android, iOS, Web, Auto, and TV were each rebuilding the same logic, and crashes kept climbing. I proposed moving that logic to the backend and making each client a thin renderer. We proved it on TV first. A full app on a new client — browse, playback, search, library, recents — had taken a team of 4–5 engineers about 12 months. I wrote and shipped the Roku, Xbox, and Fire TV apps myself, and that same launch dropped to one engineer and about 3 months, so the team could take on several clients instead of spending a year on one. As tech lead I oversaw Android TV, Apple TV, and the rest of the Fire TV work. Podcasts used that same bet to hit its launch date.

12mo→3mo
to launch a full client app
4–5→1
engineers per new client
~80%
drop in client crash surface

In the press

02 / approach

Product first. The stack is whatever the product needs.

  1. 01

    Start from the customer problem, not the architecture

    I pick projects for the customer impact they drive, not their place on the roadmap — and I'll rewrite, drop, or pivot the architecture the moment it stops serving that.

  2. 02

    Measure the thing that actually matters

    At Meta, the number that mattered was confirmed daily posts, not a better model — volume of posting is what actually moved the goal.

  3. 03

    Quality bar stays; throughput moves

    Agents changed how fast I build, not what I'm willing to ship — the discipline moved from reading every line to planning up front and validating with tests.

03 / life

Off the clock, still a person.

All work and no texture is a boring page. A little of how I actually spend attention.

  • Two kids.

    Being present at home has been the real work. Dinner table, bedtime stories, weekend mornings — those come first, no matter what else is going on.

  • Interests that stuck.

    Sunrisers Hyderabad, and Test cricket in an age that mostly doesn't watch it. RPGs: Black Myth Wukong, Red Dead Redemption 2, God of War. Indian cinema on opening day, MCU on the side. YouTube Music — eight years on Amazon Music, then I went back.

  • Cooking, and biryani above all.

    I cook. The dish is Hyderabadi biryani, vegetable and egg, any day. It's the one dish that always feels like home.

04 / contact

Email is the fastest way to reach me.

avrohith@gmail.comLinkedIn