Rohit Padmakumar (@rohitpadmakumar) header
Rohit Padmakumar (@rohitpadmakumar) profile picture

video + social @GammaApp I prev. @sandwich

Los Angeles, CA
Joined July 2016
Gamma just raised a Series B at a $2.1B valuation led by @a16z! We've also crossed over 70M users and $100M ARR profitably. Earlier last year, when I was at Sandwich, Gamma became our first client as part of a fellowship I made where we took a bet on startups by making a special commercial for a cash + equity split. It's moments like these that make those bets worth it. And since going in-house, we've just been on a tear: our rebrand, Gamma 3.0, and now this fundraise. I'm so incredibly proud of this team and all the awesome work we've done. Had a lot of fun making our announcement video. Shoutout to @malloops_ @neversitdull @eostudi0. If you're a filmmaker, creative, or editor looking for your next move, DM me. We have a lot more coming.
11
2
1
100
12,976
A storm in Lisbon made us cancel our big event we had planned in this epic castle. So I hung out with the chef for a bit to hear his story. Couldn't not make this thing a content opportunity.
5
259
JAMES IS THAT YOU??
1
3
126
Rohit Padmakumar retweeted
We built a harness that proves bigger isn’t always better. It restructured our entire platform. At Gamma we serve millions of LLM requests every day, so it’s crucial for us to get big-model quality without big-model costs. But how do you swap in small models without sacrificing quality? We broke up tasks for smaller models to conquer individually, and then tested models at large scale to measure their visual taste with real users. The result: costs dropped, quality rose, and we gained a powerful dataset for training our own models. I wrote about what we've learned deploying small models at scale at Gamma: research.gamma.app/2026/11/s…
13
8
9
52
286,184
Rohit Padmakumar retweeted
There’s no overarching “best” among smaller models. We literally ran experiments looking for it. Part of Gamma 5’s new harness deploys smaller models for specialized tasks, rather than offloading all roles to one big (and costly) model. To do this, we wanted the best smaller model for the job(s). But after running live experiments with real Gamma users, alongside weighing model costs and speed, we found that there was no single reigning champion. There was, however, a “best” for each task; that’s why Claude Haiku handles our edits, while GPT-6 Luna is entrusted with reviewing finished decks. Check out the full scope of the experiments, along with our “bests” for each role:
We built a harness that proves bigger isn’t always better. It restructured our entire platform. At Gamma we serve millions of LLM requests every day, so it’s crucial for us to get big-model quality without big-model costs. But how do you swap in small models without sacrificing quality? We broke up tasks for smaller models to conquer individually, and then tested models at large scale to measure their visual taste with real users. The result: costs dropped, quality rose, and we gained a powerful dataset for training our own models. I wrote about what we've learned deploying small models at scale at Gamma: research.gamma.app/2026/11/s…
5
1
24
1,601
Rohit Padmakumar retweeted
Gamma is proud to join the 2026 AI 100! Thank you to @NewcomerMedia @BessemerVP for the recognition. Excited to keep pushing AI forward alongside our fellow builders.
7
5
26
2,472
Rohit Padmakumar retweeted
Gamma 5 does killer image generation! Yes, you can generate images directly from the prompt box in Gamma 5 to use in your presentations or as standalone assets. We have a suite of models you can use, including GPT, Nano Banana, Flux, Ideogram, and Gemini. Ask Gamma anything, give it any reference, throw in a photo, and create some magic.
3
4
18
2,273
We're hosting a Coffee for Creatives in London next week! ​We’re back at Photobook Cafe for a cosy morning with creatives, creators, and brandmakers. Join the Gamma team to chat about taste, creativity, and AI over coffee & pastries. On us! I absolutely love these so much. COME BY! RSVP in thread.
2
1
122
Why ban PowerPoint? Many MANY reasons... New case study with our friends at @AnotherCo!
80
Cafe meetup! @GammaApp x Dust! In London! Join us in Shoreditch to chat about all things AI at work: how your team is using agents, what's working (and what isn't), and how to communicate effectively in the midst of information overload. 📍The Photobook Cafe, 4 Leonard Circus, London 🕤 Friday, November 13, from 9:30 a.m. to noon GMT Coffee is on us. And if you want to talk about Dune 3, I'm ready. RSVP in thread.
1
92
Rohit Padmakumar retweeted
Oh hello world, this is Gamma 5 if you need a little refresher.
11
3
1
42
3,025
AAAAND we got a new commercial for Gamma! Gamma is all about building a world with your best ideas. Whether you have a killer campaign to show everyone or an upcoming pitch you need to nail for your boss, Gamma is the best way to visualize your ideas fast into presentations that actually look good. We partnered with the team at @creatorcamp_ to bring this gem to life. Directed by Megan Tan.
1
9
320
New Gamma 5 video! A quick little explainer we put together to show off all you can do with this new version of our product.
1
9
241
Gamma is coming back to London this November! We're planning out a series of events with our partners and community for a week-ish. And I'm always looking to connect with more creatives in the world, so if you're around, let me know! Matcha is on me.
5
151
Rohit Padmakumar retweeted
Gamma 🤝 @CP3
3
10
2,009
In case you missed our Gamma 5 launch video, there's also this neat 30-second cutdown that just came through Slack this morning. This one's got even more energy 💥 💥 💥
1
1
12
425
Our Gamma 5 launch was our biggest video ever. When we were brainstorming how to go about this one, we wanted something a little more energetic (almost like a product sizzle) than our previous launches. I LOVE our team, so we wanted to showcase more employees than ever before, with a bunch of us in our "oner intro" and having Grant, Jean Ellen, and Jordan talk about Gamma 5 from their perspectives. Iverson absolutely nailed it with this one, with Jonathan Chekroune and Andrew Wendling bringing a lot of our creative ideas to life. Some fun rando things on set: - We didn't rehearse the intro at all and improv'd it in 45 minutes in a Berkeley parking lot - We had to reuse that one matcha for so many takes - The sets were built in our brand colors (and we broke them down after each interview) - I fell maybe 5 times I love that we get to be so fun and so creative at Gamma. @iverson
🤖 Made with AI
1
1
9
213