This is what egocentric training data looks like at the sensor level.
↳ 1920×1080 @ 30fps, head-mounted
↳ 9-axis IMU at 563Hz, hardware-synced
↳ VTS-aligned timestamps (sub-frame accuracy)
↳ Accel + gyro + mag + temp, every frame
↳ Annotation and action labeling and Depth pipeline about to integrate
If you're training manipulation, tool-use, navigation, or any physical AI model that needs first-person human motion data at scale
We're the supply side infrastructure you don't have to build.
DMs open.
Soon going to launch website as diverse Egocentric data marketplace .
A month ago I was talking to @xuster in IITK and asked him:
"Everyone is collecting robotics data. Who's actually going to win?"
His answer was simple:
Either build the highest quality datasets with the best environments and hardware, or build the most diverse datasets.
And this is why we got bullish on diversity at @Datra_ai
Most people are playing a volume game.
We're betting on diversity.
Over the last few weeks, we've been collecting data across factories, hospitals, workshops, and educational labs. But one thing we're particularly excited about is partnering with Engineering anf ITI colleges.
Why?
Because these students spend years learning practical industrial skills before entering the workforce.
Welding.
Wiring.
Electrical assembly.
Machine operation.
Workshop tasks.
These are exactly the kinds of skills robotics companies ultimately want to automate.
If future industrial robots are going to learn from humans, it makes sense to learn from skilled humans performing real tasks.
We're now collecting data across industrial environments, hospitals, and technical training labs to capture a much broader distribution of real-world human behavior.
Also looking to connect with robotics labs and teams building VLAs, world models, manipulation systems, or industrial automation. I'd love to understand what you're building and what data is actually missing today.
Been quiet for 1 months.
A lot changed.
We finally started deploying our devices across Factories MSMEs in India.
Every day we go factory to factory figuring out:
which environments produce the most valuable real-world robotics data.
So far we're working with:
textile, leather, tanneries, packaging, manufacturing, households, Engineering and Nursing College labs and workshops and hospitals/clinic labs.
The interesting part is not just collecting video.
It’s:
diversity of environments
continuous workflows
exclusivity of datasets(Because we're collecting directly not buying from vendors)
and processing everything end-to-end ourselves instead of outsourcing to vendors.
We’re building a real-world multimodal data layer for physical AI.
Scaling much faster this month with better hardware deployments.
Also looking to partner with hardware + robotics labs working on sensing, wearables, perception, or physical AI infra.
@ycombinator came to IIT Kanpur today.
Had a great discussion with @snowmaker about what we’re building at @Datra_ai along with @xuster, @agupta and @TheChowdhary
Excited about the future of physical AI and and how we can take forward This Data infrastructure.
Got into @fdotinc.
We’re working on @Datra_ai collecting and converting real-world multimodal data into training-ready datasets for robotics.
Physical AI won’t scale without better data. That’s what we’re solving.
How the dots are connecting:
At @Datra_ai , the real bottleneck isn’t demand it’s building scalable, on-ground ops to collect high-quality real-world data.
Most teams struggle here.
We don’t.
Previously I built TaskSwap(https://t.co/umZI1ROGcu) scaled across 30+ colleges in North India.
What we kept wasn’t the product, but the network.
Now we’re using that same student distribution to collect egocentric data from households and real environments.
While others are still figuring out how to build supply, we’re already activating it.
From our last post, we’ve had 2-3 angels ready to commit ~$80K.
In the last few weeks:
3 paid pilots ($3k–$5k each)
Pipeline with 5–6 leading robotics companies (billion-dollar category)
Early factory + household data collection network live
The bottleneck isn’t demand it’s scaling hardware and on-ground ops.
We’re opening the remaining allocation in our $250K angel round.
If you’re an angel interested in backing the data layer for physical AI, happy to connect.
This is Exactly the problem we're trying to solve by collecting Multimodal Data and providing it to the top Robotics labs to train their Robotics model
We're looking to rent 1,000 wearable cameras to collect egocentric data from factories.
Specs: 1080p, 30fps, IMU.
Form factor: head-mounted.
We're already in talks with 2-3 manufacturers but looking to connect with more especially from China.
If you build or supply head-mounted cameras for data collection, DM me.
This is urgent. We have demand, we need hardware.
@Datra_ai
It’s been 1 month working on egocentric data collection from Indian factories. We closed 1 partnership and running 3 pilots (including @scale_AI ).
But here’s the truth: being a data vendor is not the endgame. I don’t want @Datra_ai to be that.
We’re now moving toward owning the full stack building or importing our own wearables, collecting egocentric data with full ownership, and using it to train our own robotics models here in Bangalore.
This is harder, slower, and more capital-intensive but it’s the only way to build real leverage in physical AI.
We’re looking for strong backing from angels and early-stage VCs who believe India can build foundational robotics models, not just supply data.
one month ago, we launched cohort 7 of
@residencyBLR; since then, builders have been shipping, learning, and pushing boundaries every single day.
here's a glimpse of what’s been cooking:
@Ashf03 launched @AquinF03. 20K+ subscribers on youtube. 200+ signups. won a grant and added more fuel to his journey.
@AskBaluBabu selected Top 3 out of 700 to pitch the OpenAI team in India. in active talks with @peakxvpartners & @scaletogether. 3 pilots moving to LOIs. got an amazing co-founder onboard
@sureshmohan_ tested his inference engine with 2 AI labs in the US. incorporated and got a cofounder. Beta dropping this weekend.
Bharat and Prithvi - building @cinic_ai. ran their first hackathon in Hyderabad. got @invideoOfficial, @KoyalAI & @mosaicagg as sponsors. 54 paying customers and counting.
@niraj_munot building @maticalgos, integrated with 3 brokers. 900+ users. 2 podcasts recorded.
Shubham is building @rakshamLabs, a made in India NavIC GNSS module. flight controller nearly ready, first flight coming up. team growing and VC talks underway.
Pushkar building @Datra_ai secured a @builddotai pilot, partnered with OmGrab for wearables, and in talks with Scale AI.
Virakti and Manasvi building @ai_genera. 50+ creators onboarded, 10+ AI films ready, MVP OTT live, and a growing AI-filmmaker community.
@PossiblyTejas building @OdstatHQ, found a tech co-founder. completed beta and launching now. onboarded a new data provider. initiated brokerage API talks.
Rahul building @staaakeIN launched the visibility sheet. paid subscribers within days. Built a 60K+ Luma community.
@addddiiie building https://t.co/S0Q5kab4Te closed their first enterprise client. 6 more enterprise deals in pipeline + VC intros flowing in.
this is just half of the cohort, rest of them are just as amazing. stay tuned!
next cohort application are open. link in comments.