Our 2024

Before announcing Project 11A, I have some other related news. Aeos is now 300+ full time employees, profitable, thriving, and bootstrapped. We’ve had a period of insane rapid growth lately. Most people assume I spent most of my time in the last 2 years creating content, but that has been mostly automated for a long time. Achina (my wife & co-founder) and I have actually spent most of our time building a company in relative stealth. A majority of our revenue is B2B, and a lot of the crazy growth has been in the last 6 months or so (word of mouth of our outcomes compounded I guess). We now have a cool 25k sqft of office space in a comfy (and close to the airport) part of Bangalore.

Here is an (incomplete) list of clients we worked with in 2024-2025. This includes advertising and services. There are many more we’ve worked with but I couldn’t fit it in one image. We work with 100s of enterprises now across the Aeos ecosystem.

Aeos is structured very differently. It’s kind of like a parent company that has multiple companies with full management teams under it. I knew content would drive maximum RoI in terms of deal flow, which is why I spent most of my time there, but the subsidiaries use that dealflow. Two standout companies for us have been YAAS, which does content as a service for companies with AI assisted cost advantages and Labs (also called Aeos Labs) that does AI services. Both services companies have completely different unit economics compared to traditional services companies because of Gen AI. Aeos (the parent) itself is a lean team that runs experiments and builds distribution. All channels that we own are under the parent.

Let’s talk about some of the projects YAAS has delivered in 2024 before we get to Labs. YAAS is run by Rohit Kamath, Loveena Kamath and Ronit Mangnani.

Clearly Tripping: A channel built for Cleartrip (the travel brand).
0 to 247k followers in 7 months.

Builders Central: A no-code/low-code channel built for Zoho (A global SaaS company).

Zerodha Markets: Here we edit 1 long-form a day per channel - where their aim is to cut through the noise and simplify the biggest news in the financial markets, yet go a level deeper compared to traditional media. We also edit for this channel in multiple languages.

Zerodha Main Channel: We do some of the vids here where we cover documentary style videos around the indian stock market + a new series on scams. 

Grow Kwik: A channel built end to end for the popular eCom checkout company.
0 to 300k followers within 9 months.

BingeWealth: A finance channel built for Bitsave.
170k+ followers in just 142 posts.

We have several more established ones, but I want to feature a couple of new clients too:

NRI Shaala: A channel for NRIs in the US built for Vance Tech (A popular money transfer platform). First long-form video hit 400k views and 10k subs. Sometimes growth can be fast if the content pieces (in long form) are really good.

Lumio:
For Lumio, the first video hit 250k+ views, and one of the first 20 videos crossed 1M+ views. It’s very early days for our work with these two channels, but wanted to show you a snapshot of what the early days look like.

Bangalore City Police: Helped the police use AI avatars +edit to share important updates with the city.

Work for top cricket Teams: Used AI avatars + edit to bring players closer to fans in Kannada and other languages for some of your favorite IPL teams. Our NDA prevents us from publicly talking about the projects, but you’ve probably seen some of those videos on the internet! 

Zero1 Network: We do a lot of the groundwork like edit for this creator support network in collaboration with Zerodha, where we help creators tell their stories. Channels like Anurag Bansal, Nupur Dave, Pranay Kapoor, etc are examples in the network. Two of our own channels, Full Disclosure and Breakdown are also part. The network is spearheaded by Sagar Gudekote and Prateek Singh from LearnApp and I promise you they’re on to something special.

All of these stories are just a small fraction of our client base rn. We took our sweet time to make the first few work but the inbound dealflow over the last few months has been kind of crazy. People don’t just want content, they want us to hit the bar that we hit on our own channels and doing that requires us to choose clients wisely and build culture and process that makes creativity possible at scale. We make over 1,000+ videos a month end to end now so you can imagine the sheer scale and coordination effort to pull this off every month. 

2025 at YAAS is good so far. We just closed Vodafone-Idea and several other enterprises this month. We say no to more deals than ever.

Btw, we do not have a 100% success rate. Sometimes there’s a mismatch in expectations or some very specific thing that we can’t execute on. Sometimes it’s bandwidth issues. But in general, clients have had a very good experience with us and most of our growth is word of mouth and account expansion. I’ve only ever made one tweet about YAAS and haven’t yet posted a reel about it. For all our distribution, we haven’t actually used it for YAAS.

Labs

Now, let’s get to the other entity, Labs. Labs works at the intersection of content and AI. We started out building products like Godinabox (GPT API on Whatsapp), AlphaCTR.com (AI thumbnail tool), and Autocodepro.com (Prompt2Code tool), but eventually realised it is a better business model to use both our tools and third party tools behind the scenes, and charge for services. Executives are busy and dislike looking at some SaaS platform with a 100 options, so they’d rather we deliver end to end outcomes and use whatever tool we want as long as expectations are met. We also think the form factor for SaaS will move to a chatbox over time for this very reason. Labs is run by Harsha Reddy and Tejas Tholpadi. We also wanted a focus on a niche: AI + content (we still categorize some niches like calling/interviews under content because the differentiator is how the bot on the other end feels). Generic AI services companies aren’t doing that well because they have no idea who their customer profile is. We realized early that a focus on content is useful and fits our ecosystem better.

Here are some projects Labs has delivered, includes projects in the prototype/early use stage as well as mature ones and most importantly tools we use internally too:

VideoVault: So first off we built a NAS based collaborative solution to handle over 1,000 videos a month at Aeos. Basically our own github for video, complete with version control and timestamp based editing. But the thing is that it was important to have this work locally and on-prem. Most post-production assets like 4k podcasts shot on cinecams exceed 100GB making cloud-based collaboration slow af. Uploading and downloading these files over the internet can take hours.

With VideoVault, you can collaborate locally where read/write speeds are 10x faster than typical internet transfers. This isn’t a trivial problem. We make sure videos are permission gated, double backed up locally + a last resort backup on the cloud also useful for external sharing. Attached an arch diagram for those technically curious and a screenshot of the tool. We’re going to start giving access to VideoVault to our customers soon after an Aeos-wide rollout first. This is how we make software now: mostly for internal purposes; and if it works well, we spin it out and provide it to our other customers.

Sportskeeda - SceneStitch:

SceneStitch is an AI-powered platform designed to create short-form video content at scale. It enables users to generate posts for social media or produce hundreds of variations of a single ad for A/B testing. The platform utilizes AI to craft everything from scripts, avatars, and screenplays to audio voiceovers, and integrates third-party content repositories to enhance videos with relevant imagery.

Redbull: Longform to Shortform Video Automation 

We created a prototype and product spec for Redbull’s Global team for a tool that allows them to take a large longform video and cut it up into smaller short-form videos ideal for social media consumption.

Amazon Prime Indian Police Force (IPF) Deepfake Campaign

A fun deepfake UGC campaign that lets users step into the IPF trailer video. By uploading their name, phone number, and a selfie, users can see their face transformed into a police officer character in the video trailer. This was done in collaboration with the agency MBCS.

Amazon Prime Mirzapur S3 AI Launch Campaign

We expected 3K users to show up per day, but we saw 60K users show up each day. Again, a trailer in which you’re deepfaked in.

An interactive WhatsApp-based engagement campaign for Mirzapur Season 3. It immerses users in Guddu Bhaiya’s and Golu Didi’s world (the lead characters of the show). We did this project with MBCS. 

Rajasthan Patrika AI Community/HR Manager:

Running gated communities is tough and engagement is actually a content problem. We decided to let AI handle it. For this client we built an AI powered community manager - that sits on your favorite messaging platform (WhatsApp, Slack, Telegram, Discord). 

AI-based Screening Prototype for Godrej Properties 

We built a commissioned prototype for Godrej Properties. A tool that automates the screening stages. It generates candidate profiles from resumes, conducts AI-powered audio interviews, and scores candidates based on recruiter-defined parameters. This automation reduces manual effort, saves valuable man-hours, and streamlines the recruitment process. It still doesn’t remove the human from the loop though, so don’t worry – interviews aren’t getting automated anytime soon.

Photorealistic Virtual Tour with AI Voice Agent Prototype for Godrej Properties

We built a prototype using Gaussian Splatting to create photorealistic 3D tours of flats, accessible via PC or mobile. An integrated AI voice agent answered real-time queries about the property, amenities, and location.

I asked the team to setup one for one room of our office for you to try. You can give it a spin here: https://office-splat.vercel.app/character


Navaratri Celebrations with Santoor 

We launched an exciting Instagram campaign for women during the 9-day Navratri celebration, allowing them to create AI-generated images of themselves dancing in Garba attire. Each day, the campaign shared a unique ‘Young Soch’ message inspired by Durga Ma’s form.

Mother’s Day Celebration Campaign for Sunfeast

A Gen AI marketing campaign exclusively for the Sunfeast customers in the UAE, where participants can upload their details along with a picture of their mother. The image-model based workflow then creates a beautiful canvas-style image using predefined artistic styles. Users can then either download the personalized artwork or share it directly on social media, making it a heartfelt and creative way to celebrate their mothers.

Personalized Video Invitations for 25 Batches at IIT Madras Reunion

IIT Madras hosted an annual reunion to welcome alumni back to campus. This year was particularly special, as it brought together alumni from 25 graduating batches. Instead of the usual standard email invite, the organizers decided to make it more personal. 

Daawat Biryani Story Campaign

A personalized video campaign that uses Deepfake workflows to let users star in a video with celebrity chef Sanjeev Kapoor. 

Lenovo: Murder at Breakfast

We created a short film with Lenovo India, FCBSix & Anuja Chauhan. It was a murder mystery set in a uptown Bombay 5-star hotel. One of our first forays into AI film making. We used a combination of Flux to generate images and RunwayML to create videos from the images.

As you can see some of the stuff is hard, some are simple (but the market is willing to pay). We do not judge. There’s a lot more stuff that Labs has done but I won’t bore you. Let’s move on. 

The big goal with Labs is to keep improving our margins on creating content and automating or reducing the error rate with content related tasks. It’s built with the express intent of trying new models and offering the latest in content + AI to our customers. There’s actually a lot of skill in making a model do what you want. It’s why other people can’t make tools like Heygen do what we do with it. Labs is less geared toward scale or profit than YAAS and in many ways built to accelerate what we do. Labs and YAAS work together on several projects that involve both avatars and content like the RCB project. We also build software to try and give ourselves efficiencies and packaging it into a greater service rather than selling it for 5-10 dollars (that model is a race to zero imo and hard to break out now and you kinda get stuck at 500k-1M in ARR).

Across both companies, our radical focus is on GenAI + content. It’s unlikely we could have bootstrapped this business without GenAI helping us make the unit economics work. The biggest benefit has come from our usage of AI Avatars. Our first experiments with this started in Feb 2023 with Wav2Lip, and we faced quite some ridicule from existing content creators on my belief that people will watch AI presenters. A lot of time has passed now. I was at Davos recently, and this handy little book was given out by the YT team. 

My own channel is now recognized as one of the largest avatars in the world. You may not see it as innovative because it’s gotten easier today, but we did it at a time where nobody believed in it. At 60-100M views a month on my own account, it is probably one of the largest “wrappers” from the country. I’ve actually been trying this since early 2023 lol. Today the avatar is run end to by Sagar Amruth and managed by Shreyansh Bhuwalka.

It’s a bit weird to actually see it work and get to such scale.

Now of course, not only would we build content for our clients, it made sense to also build it for ourselves. Our own channel roster includes my channel, AevyTV, Full Disclosure, 100xengineers (helps with engineering talent for Labs), Overpowered, Breakdown, Mr.Nerf, Aevy Video School, and many more. We do both long and short form and have set a bar for quality globally on the kind of quality of editing and storytelling we do. In December of 2024, we did 400M+ English views across our own + client channels. On our own channels we cumulatively have 7 Mn subs + followers across. 2 years ago, we had <500k followers and just 1 or 2 channels. I want to emphasize again that the quality of our content/art/direction/editing and use of technology is now internationally recognized as top class. 

All of this is possible not just because of AI B-roll or avatar efficiencies, but because my wife built out a handy little video editing school that has produced over 3000 video editors over the last few years. Everyone from the WTF podcast to Ali Abdaal to Blinkit has hired from her school. But the big game changer was YAAS’s ability to hire the best editors trained in her editing style. Having 100+ very high quality editors a quarter available to hire is a serious moat. Most people do not realise the sheer demand that is bubbling up for video editors. Solid execution and love for students and art made this happen. Video editor salaries have also shot up! She will never scale this up to 10x or 100x the scale. Quality in schools is all about taking a tight small cohort and making sure the students are actually skilled at the end of it. We need them to be skilled to hire them at YAAS so the goal is not to be an edtech and mindlessly scale. Ocus, Sachin, and Martin now run the school and do an excellent job while Achina is spending most of her time now running Aeos with me. On a side note she’s probably the best co-founder I could ask for at Aeos!

We started Labs out believing we should build products, but quickly changed our mind as we learned (pretty much on day one when we used GPT3 in playground and DallE-2) that AI disrupts content creation. We realised enterprises now care more about outcomes than paying 5 or 10 dollars for a tool, and we also realized many of those tools in the 2010-20 era did well not because of exceptional engineering but because they had gamed SEO really well. We also recognized early that standard SEO would die after LLM use, but people will still search and discover new products via YouTube and Instagram. Luckily I’ve had experience building SaaS platforms (we did the whole hog with my last company) and realizing first hand that it’s more about distribution than product, so we didn’t repeat the same mistake twice. 

At the rate we’re going we may as well be 1000+ employees in the next 18 months or so. We are hiring 30-50 people a month. I’m hoping video technology gets better and we do not need to have to have so many people. Culture is actually better the smaller you are, so I push harder on staying small. Despite being 300+ people, I am still a believer in small teams. A company’s arteries stiffen at scale. And scale is also problematic in some ways – the office coffee bill is now 1.5L a month. We’re bootstrapped so our margin for error is basically 0 and has been for 2 years.

But we’ve thought about efficiency and “how can we automate this?” from day one. Even the backgrounds in our videos are propped up on a giant LED screen with the image powered by Midjourney - we have almost no colorists/color graders in the company because of this. There’s genuinely a lot of stuff we do here that no other company in the world does.

But there’s a clear opportunity here to really help a ton of businesses create content and use GenAI efficiently. My second life as a content creator allows me to gain early access to new models very early, which seems to give us a big advantage over anyone else attempting this. Big advantage in always knowing in advance what’s happening so we can react before the market can.

The reason I didn’t reveal any of this so far was because I wasn’t sure. I’ve been an entrepreneur since I was 19. I’m 32 now. It’s been a long journey of several failures and some successes, and one can never be sure of success. The main thing is this recognition that there is no such thing as permanent success, we can all still always go down to 0. Most first time entrepreneurs don’t recognize this deeply enough I feel but I feel it in my bones. So I tread the waters with some secrecy and a lot of humility. But at this point not talking about Aeos is really slowing down hiring (our hiring team is 8 people) and I think the benefits of talking about it outweigh the negatives. 

The negatives are mainly increasing competition (which I don’t mind) but more importantly the triggering of Girard’s mimetic desires. Which is that even if someone wasn’t doing what we were doing, now it somehow upsets them and they want to do the same thing in the same manner. Many of these people will end up disliking me without spending time to think why (competition does something to the brain). I think a lot of why some of what we did worked is for the first time in my life Achina and I chose non-competitive markets with little to no competition (Peter Thiel subscriber) and did things in our own unique way. People also really underestimate the amount of experimentation we did in our early days. You’ll have to go back and fiddle with CUDA versions or shoot videos on your phone with no subscribers. I think there is a real moat in doing things your own style, especially as taste and execution becomes more valuable than skills. We also think people are very myopic of how the next thing would look. They extrapolate some old thing: people believe that in AI you’re either going to do research or build products (and charge a subscription) using other people’s models but there’s very little business model innovation because they’re not good at envisioning newer ways to do commerce. Subscriptions are a nuisance nowadays (I cancel right after I subscribe to anything) and research is roulette. Projects with high visibility and visual output like ours are best done as a service with a combination of models and people at the backend. You’ll probably come around to this conclusion soon too.

Unlike others, I’m also very bullish on all media companies with distribution provided they know how to use it. We haven’t attempted to win any of the text/writing content markets, but I think there is a massive massive opportunity there which we won’t go after for focus reasons. In fact, I expect a role/status inversion between tech and media shortly as tech totally becomes a commodity and attention becomes more important than ever. There’s too many repeats of the same kinds of products advertising with us. This is going to sound funny but I can actually sense saturation of a space from brand dealflow lol. The big existing products will get bigger though but it’ll be hard for new entrants to break out.

Lastly, I want to say that my role at Aeos is idiot.

Calling myself a content creator/idiot was the most liberating thing I ever did, free from having to shape myself into some existing “role” or “field” and allowing us to see everything in a new way. Calling yourself a content creator, which isn’t taken seriously yet by anyone “important”, was a way to get me totally out of the status game. To get me out of the game of “oh I need to build this thing from scratch” (nobody really builds anything from scratch, we don’t write assembly and even if we did, we don’t know how to make computers), to allow me to truly be something that didn’t fit. I urge every misfit to call themselves an idiot or content creator, it helps you break out of monkey status mode. Social media is built on status and you should never conflate that with seeking market truths because the crowd is usually wrong. In fact, there’s a huge (and widening rift) between private and public information. Almost everyone in the know knew that things were going well for us in the last year or so, but we simply spoke about it online. 

Also, we broke a lot of rules. We didn’t have a logo until maybe a couple of months ago. YAAS still doesn’t have one. We don’t have a website for the main Aeos entity, it redirects to the Labs website. Our subsidiary thing isn’t properly structured yet. We built two services entities instead of one because we slowly recognized there were two different customer profiles who would make the final decision on AI + content services and content. Everything here mostly runs on trust, market truths, a lot of failed experiments, and a bunch of people who really got fed up with the regular venture funded startup game and were okay taking their time to grow bootstrapped. Valuations are a type of status game. I hope more founders start feeling the same way about dilution. At the Aeos leadership level we call out anyone playing status games. 

I have one last bullet left in the chamber, Project 11A.

Once you do a little well and you have some free cash flow, that is actually the time to push the boundary and innovate. We will do that now and show you that priorities should always be to make something work commercially first and then innovate once it does. I had it the wrong way in my head most of my life like every other engineer. Most engineer-entrepreneurs have it the other way around too. Actually it would kinda be impossible for me to build this type of company without first going through the journey I’ve been through. Prior failure is important.

In this company money is meant to try new things, we don’t pride ourselves on revenue or how much profit we make. It is meant to be spent on doing fun stuff I’d always want to do as a kid. I cannot understand people who run companies with purely financial goals.

Since I’m not CEO of any of the subsidiaries and day to day is run well by the teams, I have the time to try radical experiments at the Aeos level. Let’s see what that ends up being. If you come to our office now, you will recognize that we are good executors and certainly took some bets that didn’t seem clear back then/looked distracted but worked out now. But if you believe we are not innovators, then you may change your mind after 11A. We think India can be a global powerhouse in this adjacent industry we’re going after.

2025 is also the year I really focus on my health. Things are moving smoothly at the cos, so I finally have the time to focus on my health, which I’ve basically ignored for a decade or so. Not anymore. Bootstrapping is really hard, but once it sort of works, it’s a lot more chill than constantly having the expectations of building a billion dollar biz, so I can breathe a lot more freely and focus on repairing the damage I’ve done to myself lol.