project screenshot


(i) decentralised data store for nano-satellite industry + (ii) compute to securley data to train geo predictive models on sensistive data + (iii) marketplace - built using ocean procol + secret network


Created At

Scaling Ethereum

Project Description

nano satellite industry is blowing up!

every year it gets cheaper to launch a (ever smaller) nano-satellite. some companies are launching constellations of up hundreds of these table sized boxes with a variety of payloads or receivers collecting a wealth of information - from weather data, to ais, to radio frequencies, to photographic parameters.

this data is quite sensitive by nature so these companies are reluctant to sell it, except to the whales. lets decentralize this industry and make this data available to ml developers who can build innovative and integrative solutions and deliver them to end consumers through a marketplace!

nanosat startups need revenue and have these boxes floating in space delivering data, it doesn't matter if they have customer paying for it, the data is coming in all the same! they will be happy to make extra bucks for posting their data for machines tot rain models. think about the power of giving access to everyone GPS location to a predictive model! what great analytics we can develop, and at what great accessible price, its not only the hedge funds and governments that should have and trade on this valuable info!

How it's Made

  1. data store - created using aws + ipfs + ocean + secret - idea is for data providers to be able to host their data on chain or on premise with the guarantee that their data will be totally secure and only use for "computer-to-data" purposes and such purposes will be tracked and the models trained on this data will also stay on-chain as to prevent data leakage.

  2. compute to data - using ocean and secret network, docker and pytorch, provide a toolkit for ml researchers to be able to build and train models on chain (who's weights can never leave the chain) without actual developers being able to see the data these models train on (ok, they can see headings or example data provided by data provider/vendor). it is also interesting to allow developers to build in pipelines that will allow end customers (in 3-marketplace) to be able to customize this data (i.e. i want to train this model to predict traffic flow in a specific country, the model might be pre-trained with info from the whole world, but i want it to get smart on a specific city, so i can request a re-training using specific data coordinates).

  3. marketplace! regular joe should be able to access predictive analytics from models for the price of a cup of coffee! they should also be able customize their solution before and after purchase, put up bounties for analysis they want, and many more features. once customer pays, some % goes to system and rest gets split between developer that made model and providers that provided data to train models (auto tracked base don value of data per mb or some metric).

background image mobile

Join the mailing list

Get the latest news and updates