How NATIX plans to disrupt autonomous mapping with decentralized data

NATIX Community has formally launched its 360Data Subnet (Subnet 72) on the Bittensor decentralized AI community, representing a step ahead in its mission to modernize mapping and autonomous driving by way of decentralized, real-world knowledge processing.

Powered by a Solana- (SOL) based mostly decentralized bodily infrastructure community of sensible cameras, NATIX combines high-volume, street-level knowledge assortment with scalable machine studying by way of Bittensor (TAO), crypto.information can solely report.

The subnet, incubated by Bittensor infrastructure accomplice Yuma, consumes 360° video feeds captured from NATIX-enabled Tesla autos and different cellular units. It then processes the info into AI fashions that enhance real-time map-making and automobile autonomy.

The transfer is per developments seen throughout the autonomous mobility trade. Notably, corporations like Uber are investing in artificial simulation and bespoke knowledge collections to raised handle their self-driving applied sciences. Nevertheless, that is costly and lags behind real-world modifications.

To handle this, NATIX is eyeing a worldwide group of over 250,000 drivers who collectively logged over 170 million kilometers by way of smartphone and VX360 system utilization.

VX360 and decentralized knowledge coaching

The VX360, developed in collaboration with Seize, leverages Tesla’s present digital camera programs to gather 360° imagery with out requiring costly new {hardware}. The information is processed each within the cloud and on the edge smartphones detecting components like site visitors lights and indicators in actual time. Extra intensive classification is dealt with off-device.

Via Bittensor’s decentralized framework, NATIX miners are rewarded for coaching and enhancing AI fashions on the subnet, that are then redeployed throughout the NATIX Edge Community.

“The NATIX 360Data Subnet on Bittensor represents the convergence of Bodily AI and decentralized intelligence,” mentioned NATIX CEO and co-founder Alireza Ghods. “By decentralizing knowledge evaluation, we allow the continual enchancment of AI fashions, considerably enhancing mapping accuracy, autonomous automobile security, and real-world responsiveness.”

The subnet’s first focus is roadwork detection, a crucial software for each mapping platforms and autonomous automobile navigation. Over the longer-term, NATIX hopes to increase its operations to incorporate pothole detection and infrastructure evaluation. Ultimately, NATIX expects to offer a full situation classification to assist autonomous automobile coaching.

Q&A with NATIX CEO Alireza Ghods

To raised perceive the influence of NATIX’s newest launch, crypto.information spoke with Alireza Ghods, CEO and co-founder of NATIX. The dialog focuses on the corporate’s method to scaling decentralized knowledge seize, why it believes its knowledge can compete with tech giants, and the way it plans to create actual demand for its token economic system.

Your entire Q&A is beneath:

crypto.information: Tech giants like Waymo, Tesla, and others and even Mobileye have spent years and billions of {dollars} accumulating driving knowledge and growing autonomous tech. What makes NATIX imagine it may supply knowledge or AI providers at a comparable high quality? Crowdsourced knowledge is nice for scale, however can crowd-collected digital camera footage actually match the precision of devoted mapping autos?

Alireza Ghods: Information assortment for mapping and autonomous driving is tremendous costly. Massive corporations that gather this footage are usually not sharing it as they use that knowledge to have an edge on the product and providing. None of those gamers have been accumulating knowledge on the scale that NATIX can do in a crowdsourced method. The most important open-source dataset for autonomous driving is Studying to Drive (L2D), a dataset inbuilt collaboration between two tasks collected with equivalent sensor suites put in on 60 EVs operated by driving colleges in solely German cities over the span of three years, which resulted in solely 5k hours logged. For the reason that launch of the VX360 on Might 2nd, 2025, we have now already collected 2k hours of driving in solely 10 days, and with just one/3 of the full pre-ordered units reside solely within the US.

With regards to the precision in mapping, NATIX might not have the precision and high quality of devoted mapping autos, however the sheer measurement of the community and knowledge collected compensates for it, making the tip outcome nearly as good as knowledge collected by devoted mapping autos. Take Seize, for instance. Seize constructed their entire mapping answer based mostly on crowdsourced efforts, leading to maps extra correct than Google Maps in SE Asia. For crowdsourced efforts to succeed, you want a big community of drivers/contributors. With over 250K registered drivers, NATIX does have such a community, and it’s the most important digital camera DePIN community globally.

CN: Gathering 360° video from probably 1000’s (if no more) of vehicles and telephones in actual time screams of a large technical problem. How is NATIX dealing with knowledge ingestion and processing at scale?

AG: For knowledge collected with the VX360, the compute we carry out for ScenGen is sort of heavy and can’t be finished on the Edge. We do the processing on the cloud stage. The information shared by customers is simply uploaded after they’re linked to their house Wi-Fi connection.

With that mentioned, we use our smartphone community, which has edge compute energy, to detect map attributes resembling site visitors indicators and site visitors lights in real-time.

CN: Are you able to describe the scaling technique: how will you go from a pilot with just a few hundred Teslas or telephones to a deployment with, say, 100k+ units streaming knowledge, with out the system breaking or prices skyrocketing?

AG: The VX360 system is sort of cost-effective. With a 350$ system, we faucet into 360° street-level imagery, whereas additionally offering utility for the drivers. Our technique is tapping into commodity {hardware}. Everybody owns a smartphone, and Tesla autos already come outfitted with 360° cameras, so it’s only a matter of tapping into what’s already on the market. If we needed to construct a 360° digital camera, it could be 20-100x dearer than VX360 by way of {hardware} price, and it could not be scalable as a crowd-sourced answer.

As for additional scaling, we’re additionally in dialogue with companions who need to function a fleet of VX360s. This could be fleet house owners or integrators who need to sponsor the price of the system for reward sharing. For example, we’re in dialogue with a Tesla fleet proprietor with over 3,000 Tesla autos of their community. We’re additionally within the means of constructing fleet administration functionalities for the fleet operators to extend the utility of our product and transcend crypto rewards.

CN: Following up from the above query, do you see these huge gamers as prospects of NATIX’s knowledge (i.e. you’ll promote them supplemental knowledge) or as opponents you might want to outperform? If it’s the latter, the place particularly do you assume your decentralized method offers you an edge in opposition to far better-funded operations with full management of their fleets?

AG: Now we have Seize as our buyer, and we’re within the means of closing just a few of the biggest autonomous driving gamers, each as direct prospects for the info and companions for constructing numerous merchandise for autonomous driving.

Each huge participant is a possible buyer, if not now, then sooner or later. A number of the greater gamers have invested thousands and thousands into accumulating their very own datasets, which quickly will turn out to be outdated because the roads are consistently altering. Even when some inform us they don’t want our knowledge right this moment, they are going to want it quickly sufficient.

Nonetheless, the AD corporations which have invested in their very own datasets are usually not bigger than 20% of the market. The remainder have both not spent a lot capital to gather knowledge and can want NATIX’s knowledge, or they’re at the moment counting on their prospects’ knowledge (if any), however can not use it to enhance the standard of their fashions for different prospects. It is a actual bottleneck for these corporations to create a holistic AD stack that may be monetized for all automotive shoppers.

CN: NATIX isn’t alone in making an attempt to decentralize bodily infrastructure. We’ve seen tasks like Hivemapper dominate headlines a number of years again or DIMO (automobile knowledge) try comparable crowdsourced fashions. Their street to success (hehe, pun meant) has been tough with restricted person uptake, token worth collapse, and many others. What classes has NATIX’s discovered or what method are you taking that differs from people who struggled?

AG: There are additionally profitable fashions like Geodnet. I imagine the winner right here is the one which focuses on protocol income and has a correct mechanism for worth accrual to the token (e.g. buyback and burn). Many of those tasks endure from a scarcity of Protocol income or poor tokenomics (lack of worth accrual). On this case, NATIX is nearer to Geodnet and is following the identical technique.

Moreover, you must have a look at every enterprise mannequin individually. For instance, Hivemapper’s front-facing dashcams gather high-quality street-level imagery for mapping, however NATIX’s 360° knowledge can cater not solely to mapping but additionally to autonomous driving and bodily AI. 360° video knowledge is the holy grail of street-level visible knowledge, and never solely is it price 5-10x greater than Hivemapper’s knowledge, nevertheless it opens the door to numerous new mapmaking use instances (e.g., POI detection with facet cameras).

Furthermore, the 360° knowledge collected can be utilized for Bodily AI, which is a a lot greater subject (ever since NVIDIA’s CEO Jensen Huang’s huge announcement on the stage at CES 2025). NATIX presents knowledge range appropriate for the coaching, testing, and validation of the autonomous driving stack and Bodily AI at a fraction of the price of centralized options that corporations like Nvidia and Uber invested thousands and thousands to develop. This knowledge is a painkiller for the simulation-to-reality and reality-to-simulation course of in Bodily AI growth.

That’s the reason we imagine we’re fixing a a lot greater drawback, which the trade is able to pay for TODAY.. Be aware that vehicles function with a surrounding digital camera system; due to this fact, coaching, testing, and validation of the autonomous driving stack is simply attainable with 360° knowledge and can’t be finished with front-facing dashcam knowledge.

CN: Past being a reward mechanism, what’s the precise utility of the $NATIX token inside this ecosystem? Will knowledge purchasers or AI builders want to make use of $NATIX or TAO tokens to entry StreetVision’s providers? If not, what’s going to drive demand for $NATIX?

AG: These are two separate ecosystems with totally different performance. NATIX is specializing in knowledge curation, whereas the StreetVision Subnet focuses on (sure) perception extraction and AI mannequin creation.

Every of those ecosystems generates its personal worth and has its personal worth accrual mechanism for his or her token. If anybody is within the NATIX knowledge, the worth accrual goes in the direction of the $NATIX token by way of the Protocol Income and buyback and burn mechanism. For these within the insights or AI fashions generated by the StreetVision Subnet, the income generated might be used for worth accrual in the direction of $dTAO (and partially in the direction of $NATIX as the info that was used for such perception and/or fashions belongs is supplied by NATIX).

Moreover, $NATIX is used for the protocol governance and securing the community. That is finished through our staking platform, which allows customers to take part in protocol governance. Validators within the StreetVision Subnet are additionally required to stake $NATIX.

CN: Your strategic partnership with Yuma provides a stage of credibility, nevertheless it additionally raises centralization considerations. DCG has really invested in NATIX, and Yuma has been instrumental in launching Bittensor subnets. How do you make sure that neither Yuma nor different heavyweight backers will dominate Subnet 72’s route or rewards?

AG: Anybody can run a validator for our subnet. Certainly, we’re working with two different main Bittensor validator gamers that may even run validator nodes. Additionally, we eliminated staking or $dTAO holding necessities for miners as we needed to make sure an open ecosystem. As you already know, miner centralization is even a much bigger concern than validator because the output of the community will depend on them.

CN: What present use instances are you able to level to for the StreetVision Subnet? For instance, is Seize already consuming NATIX’s knowledge to replace their maps in Southeast Asia proper now, or is that simply an MOU for future collaboration?

AG: StreetVision’s first job might be real-time roadwork detection. Detecting and classifying roadwork in real-time is essential for mapping updates and AV reliability. Extra use instances added later will embrace detecting potholes, street indicators, littering, and infrastructure monitoring. StreetVision Subnet may even later embrace evaluation and classification of driving movies into situations from routine site visitors circumstances to uncommon “edge instances,” as situation classification is tremendous essential for situation era for autonomous driving and Bodily AI.

Seize is at the moment utilizing NATIX’s knowledge to construct their pipelines for the US and EU markets, and they’re a paying buyer already.

Because the VX360 simply went reside on Might 2nd, we want some preliminary knowledge to construct pattern datasets, that are required by autonomous driving prospects. We’re in business negotiations with dozens of autonomous driving corporations. We’re additionally working with just a few of the highest autonomous driving analysis labs to create cutting-edge merchandise for simulation-to-reality and reality-to-simulation, which we are going to announce within the coming months.