GraphGrail Ai marketplace

0.00 ETH


1,000 ETH

Soft cap (max 0)




Launch Date


GraphGrail AI is the world’s first Artificial Intelligence platform for Blockchain, built on top of a natural language processing technology platform and a DApp marketplace.

Our mission: the creation of a strong AI (Artificial general intelligence) that will be open to all, controlled and trained by developers throughout the entire world. The concept is similar to Elon Musk's OpenAI, but for meaning.

We are developing a platform for analyzing large amounts of text data, solving the problems of extracting knowledge (IR – information retrieval) and complex semantic classification based on machine learning, neural networks and deep learning technologies. Our priorities are the banking sector, biotechnology and medicine, security and law enforcement.

The key element of the platform is a universal constructor for complex classifications of text, i.e. an AI designer.

Our team has spent the past 2 years on R&D in the field of natural language analysis (NLP), information retrieval and the training of artificial neural networks. The result was the AI designer, which allows any user to create models and train a neural network in a variety of tasks. Possible applications include but aren’t limited to spam detection, distinguishing text styles, searching for fakes based on language attributes, and testing complex conditions in smart contracts.

An important feature of the platform is that it provides a full cycle of work with data, from collection through markup and the final results. Business executives, startup owners, developers, and data scientists will all be able to utilize a rich API and the ability to create custom applications to integrate the AI designer into their own services and apps. We allow other companies to use our service and apply it in their own business.

All users, even those without special training, can earn money using the platform through creating, improving and voting on language models. The project does this through its associated GAI (Graphgrail + Ai) tokens, which are standard ERC20 tokens on the Ethereum blockchain. The models may have different attributes: they can be complex learning classifications, prediction models or simply improve the workflow for training the AI.

For the non-tech savvy, a linguistic model is simply a file ranging from 15 MB to 1.5 GB in size. This may be the result of training the AI to output certain data or a set of ontological handlers for the specific domain.

Welcome to the Blue Ocean – the AI ecosystem that allows the business to continually innovate, develop and enter new markets.



GraphGrail’s AI designer is a unified analytics platform that allows businesses and government agencies to solve the problem of analyzing and classifying large volumes of text data. The designer provides the end product and also has a user-friendly API. Today, the analytics industry is experiencing a period of rapid growth among customers interested in analyzing text.

Such customers include banks, telecom companies, retail businesses and government ministries, along with electoral commissions and political parties. All of these clients face a variety of problems: unwanted text needs to be classified (into categories such as spam, fraud, etc.) and knowledge needs to be extracted (i.e., product and service analytics).

Our designer provides a variety of linguistic models, allowing clients to quickly create an ontology (model). Programming skills are not necessary. The model can be tested and refined on the fly, supplementing it and improving its accuracy and completeness. The main advantage of the service is its ability to guarantee the client a full cycle of data gathering and analysis: it can gather the data on its own, the AI neural network can learn to analyze it, and the end result can be returned via a web interface or properly formatted reports.



Today, there is no full service AI similar to ours on the market; no one is offering an AI that can create similar models without programming. This kind of task is usually done by hiring programmers for a 5-12 month project, which only covers one specific task.

Why so long?

The fundamental problem of the text analytics, machine learning and AI marketplace is not enough data. Let’s take a step by step look at how a business may solve a similar problem (for example, analyzing its own products and services, then conducting competitor analysis):
1. Many fields don’t have a public database for this kind of data; it must be collected, cleaned up and formatted correctly.
2. Even if the data exists, there usually isn’t enough of it; AI deep learning requires from several thousand to millions of examples.
3. The data is not formatted properly; in order to use AI algorithms, tens of man hours are required to format it first.
4. Once everything is ready, the AI neural network may be trained, but only once.
5. The data quickly becomes stale; it must be continually collected and fed into the AI, which requires repeating steps 1-4.
6. Throughout this cycle, the business must hire and pay specialists, who usually have a high price tag and work only when the model must be refined.



21989a31600adb2.pngPicture 1: GraphGrail AI’s architecture

The service has two modules: a text/data analytics module and an AI deep learning module. The analytics module architecture contains three levels of data analysis: a morphological, syntax and high level analysis based on our previous development. The neural network module contains several methods and ways to train the neural network (RNN, LSTM, etc.) on the data previously formatted by the analytics module.

Thanks to this modular text processing architecture, the service allows:
1. A shorter development time of linguistic models for a given task;
2. Accumulating and incrementally improving the quality of results coming from the trained neural network, along with making its future retraining easier
3. Identify complex compound linguistic attributes (features) that are related to a particular subject area. Thus, the system implements our data processing method and outputs a new qualitative result – analyzing and classifying complex semantic data. Modern networks require large data sets to train themselves successfully; they do not know how to understand something from only a few examples.

This makes it difficult to use them in areas where large data sets have not been created, even though a couple of examples are often enough for a person to make a deep generalization. The reason is that training an AI with overly small sample sizes invites overfitting; when there are few samples and the subject area is complex (i.e. there are lots of parameters), the network will ‘remember’ special cases instead of generalizations. As a result, the AI will show good results on the data sample, but not on a real world test, as the test will be full of other special cases. The more parameters, the more training samples are required to avoid the problem. There are three approaches to solving this problem: regularization, architectural tricks, and transfer learning. The technical and scientific solutions that will be implemented in our product will be able to effectively solve the task using internal transfer learning with pre-mediated models and semantic categories. We believe the problem of “one-shot learning” described above can be successfully solved.



The GraphGrail AI platform is a meta-service that can dramatically hasten the creation of new companies and startups in the analytics sector.

Use cases:
1)The service is helpful to all blockchain companies. It can be used for smart contracts in order to check whether an external condition has been fulfilled. Such conditions can be virtually anything tied to real world objects, people, companies and their interactions: • An externally reported meeting between businessmen and politicians; • A contractual agreement; • The purchase or sale of a company; • A judicial decision; • Tracking a supply chain; 
2) The service ensures security when executing contracts. The service allows a smart contract to account for force majeure (acts of God) and similarly exceptional cases. Using an outside trusted source, it is possible for the AI to automatically check for such cases and analyze the raw information to potentially alter the contract. For example: two companies have entered into a supply agreement. However, the external conditions have changed significantly: weather conditions (strong wind, low temperature, humidity, etc.), political shifts (abolition or adoption of a law) or outside factors (a third party supplier could not deliver the components) have altered the deal. In this scenario, the GraphGrail AI platform solves the problem of providing data to a smart contract by accessing external sources such media, social networks, Wikipedia, etc., collects and analyzes data important to the contract, and decides whether the contract should still be executed. In other words, the AI becomes the procedure that provides the foundation for the execution of transactions.



Today, the service can be used for business and government analytics purposes. Current uses cases include product and supply chain issues and finding fake news. In the near future, traditional businesses will begin to migrate to blockchain en masse.

GraphGrail will be ready to provide the same kinds of analytics services that businesses use today, such as: • Speech recognition services. Audio is translated into text, then analyzed. • Bank services. GraphGrail can be used to detect credit card fraud, fraudulent access to bank accounts or personal data, and to look through third party forum posts by bad actors offering those services on the darknet. The system itself can find these posts without a human having to read them first. • State structures.  • Politicians. • Fake news. • Fraud. GraphGrail Ai is able to adapt to changing business conditions and to identify reports or complaints about fraud in a variety of industries, including new ones that do not have historical data. • Automated support systems.



According to Forbes, the analytics market will grow enormously over the next several years [2]. IDC indicates that worldwide spending on analytics will grow to 187 billion dollars by 2019, growing by over 50% in five years. By 2020, predictive analytics will attract over 40% of all investing into business solutions. Wikibon claims that Big Data spending will grow from 18.3 billion in 2014 to 92.2 billion in 2026, with a yearly CAGR growth of 14.4%. 

9fa7932b221d138.jpegPicture 3. The Big Data sector will grow by 14.4% a year



The company will use the SaaS (software as a service) business model. Users will be charged monthly, semi-annual or annual subscriptions. We will have flexible fees that depend on the amount of data processed per month, the analysis difficulty, the number of data faucets etc. We will also charge separately for API access.



Token has a role of internal currency in our ecosystem. Business client can use token to order NLP product (crowd labeling of data and custom workflow for business purposes) and ability to get it faster.

Token payout goes to people, who do crowd labeling work. Also, tokens are paid out to testers and community, who control quality of models. Balancing of demand and supply is achieved by flexible pricing - the more complex data processing ordered, the more tokens you get.



Token grows continually with the GraphGrailAi ecosystem - more clients attracts more labelers.

To be able to access GraphGrailAi Platform business must buy from 5000 to 10000 GAI tokens on Exchange. This is the way the token grows: total supply is limited.

These tokens you can spend in ecosystem: collecting, cleaning, labeling, neural network training etc.

With the Marketplace number of custom solution grows we make token growth sustainable and based on the real-life economy.



1. We use vector pretrained models - word2vec and doc2vec. These models make it possible to automatically define the semantic proximity of words, i.e. synonyms and antonyms.
2. Latent Dirichlet allocation (LDA) - allows GraphGrail to determine the relationship (similarity) between the topics of two documents.
3. Our custom development: a methodology of designing linguistic models through ontology, allowing GraphGrail to extract structured knowledge. We use a wide set of relationships, including taxonomy (category-subcategory), part-whole, and so on.
4. Specialized software tools for training artificial neural networks (RNN, LSTM) with the possibility of saving the resulting model in a file: gensim (for thematic modeling) and scikit-learn (a package for scientific data processing and machine learning).



In the process of studying the market for data analytics, machine learning, and the existing practices of using neural networks, we found a key bottleneck: to train a neural network requires a) between tens of thousands and millions of text samples, depending on the task, and b) this data must be qualitatively marked out (developers must create training samples and test sets). 

Advantages of using the platform for users include:
1. The platform is not just for brand analytics; it can be adapted to many tasks suitable for any industry, such as fraud detection, illegal activity, fake news, manipulation of opinion, and so on.
2. The AI designer allows users to customize the analysis to suit themselves, which raises the quality and reduces the processing time for their analytics. We use our own designer, which takes into account the morphology of the Russian language (or any other language) [English is scheduled for Q4 of this year], lemmas, bigrams, and complex reconciliations (by gender, number, and case).
3. An API – we will build a programming interface for those who do not want to use the web interface or who have specific tasks. This approach is the foundation for the development of a strong Artificial Intelligence: special functionality will be available that will apply various models in parallel.



Tokens will be awarded for creation, improvement, and testing of all linguistics models. Because users will be creating and sharing them, they’ll need a public database that is protected against malicious actors. This is an advantage to using blockchain, as opposed to private closed source development; the work will be sped up. Instead of a regular 5 programmer data science team, hundreds of thousands of developers from across the world will be involved. Integration with blockchain media companies will allow the service to query the blockchain and grab data for training the AI neural net. The DApp marketplace will also allow developers the freedom to test competing models, to merge them together and to obtain new (better) solutions.

The GraphGrail platform will act as the service provider for end users and developers and allows users to earn a share of the money taken in from API access sales. However, all intellectual property created by the platform will be owned by GraphGrail.



July 2017:
- GraphGrail AI ICO presale campaign

August 2017:
- Full scale testing of the product on pilot projects
- Ver. 1.0 release preparation
- Marketing campaign and ICO pre-seed round preparation

September, 2017:
- Public launch of ver. 1.0 (alpha)
- GraphGrail AI pre-ICO

October-December 2017:
- GraphGrail AI ICO
- Full launch of the platform and of the linguistic model library; launch of API access - First phase of English language support (the current AI is being trained on Russian)
- Load testing of the platform
- Linguistic model marketplace launch

Q1 2018: - Partnerships with top blockchain companies (Status, SONM, Tierion, Oraclize)
- Testing and launch of a 2nd platform language (English)
- Scaling
- Premium accounts
- Launch of the adaptive delegate rewards formula
- Patents and legal work



Victor Nosko: CEO, founder, AI, data scientist. Higher education – Southern Federal University. Winner of various university competitions, winner of Startup-Sabantui, a resident and graduate of the Southern IT Park (Rostov-on-Don, Russia). The founder of GraphGrail. Victor is a Python developer who understands the Django framework. He also understands the stack of technologies for language processing: NLTK + Celery + Pymorphy2 + GLRparser, etc. Experience with Google TensorFlow. A specialist in information retrieval, engineering of linguistic attributes, and vector models (doc2vec, word2vec). Developed a service for a complex semantic classification of large texts.

Sergey Litvinov: cofounder, developer. Higher education – Southern Federal University. Winner of the competition named after the academician I.I. Vorovych. Winner of the SCAGS Intramural Competition for Young Researchers. A resident and graduate of the Southern IT Park (Rostov-on-Don, Russia). Cofounded the GraphGrail project. He is also a designer and a system analyst. He is a Python developer who understands the Django framework.

Alexander Borodich: Venture investor, CMO. Alex is a futurist, angel investor, and a serial entrepreneur. He is the founder of the elite club VentureClub. Previous projects include MyWishBoard, MyDreamBoard, and SuperFolder. His angel investments include Future Action, where he is the Chief Dreams Officer and controlling partner. Founder of FutureLabs Future Laboratory and the crowd investment platform Alex is an investor in more than 70 projects. Alex was previously the CMO of the Group and Acronis. 

Zachar Ponimash: AI, data scientist, developer. Zachar is a specialist in the development of strong artificial intelligence. He is a C# developer with prior projects in machine learning, neural networks, mathematics, evolutionary computing, biology and physics. Zachar has experience in developing an AI framework. He has previously worked on chat bots, signal processing, computer vision, and text analysis systems, including logical inference modules. Zachar has scientific publications in peer-reviewed journals.

Dmitry Strelnikov: Blockchain and Fullstack developer. Dmitry is a specialist in information technology, Internet, and telecom sectors. Full-stack developer. Wrote the REST API for mobile applications (Node.js); created admin panels. Development experience includes a new file-image microservice, custom clustering of the map with animation. Along with a team, Dmitry rewrote most of the front end part of the portal. Full stack experience in CSS3, HTML5, JavaScript, adaptive layout, cross-browser layout, Git, Node.js, Webpack, AngularJS, React, and MongoDB.

Ilya Bredikhin: Mathematician, algorithms. Higher education: Faculty of Mathematics, Mechanics and Computer Science SFU. A specialist in differential equations. Specialization: differential equations of convolution type in functions with exponential behavior at infinity. Ilya has 4 years of experience in teaching and tutoring.

Read it in English and Russian:

WhitePaper (English) | WhitePaper (Russian)

Follow us

Site | Facebook | Twitter | BitcoinTalk | Telegram

Thank you for your time. We believe you can be our investor or you can help us grow!

Campaign author
More about the author: