Competitions

A new way to build AI models

Competition Platform

Innovative solutions often emerge from competitive environments. bitgrit's Competition Platform is designed to harness this principle, enabling real-world data skill assessments and engaging both internal teams and external expert communities with pressing business data challenges. Collaborating with partners, bitgrit offers bespoke, white-labeled or community driven data science competitions.
These competitions empower businesses to:
  • Identify pivotal AI applications within their operations;
  • Source AI models from a vast community of over 30,000 data scientists; and
  • Seamlessly integrate these solutions without the complexities of establishing an in-house data team.
For every challenge presented, a multitude of data scientists converge to design optimal AI models tailored to address specific business issues. bitgrit meticulously selects and delivers the most effective solutions from these endeavors.
Our Crowd-based AI as a Service (CAIaaS) model ensures clients receive the crème de la crème from thousands of submissions, facilitating swift scalability at a fraction of the cost of traditional team recruitment. Top-tier models are subsequently listed on our AI Marketplace, allowing other businesses with analogous challenges to benefit, while simultaneously generating revenue for the contributing data scientists.
In summary, bitgrit's CAIaaS is the epitome of efficiency in custom AI model development.
There are two competition models:
  1. 1.
    Small Business and Enterprise Competitions
  2. 2.
    Community Open Source Competitions

Small Business and Enterprise Competitions

These competitions will operate on fiat rails only, where companies pay the prize money to the winning data scientists in USD in exchange for the winning AI models, of which bitgrit will receive a fee as well.

Community Open Source Competitions

These competitions will operate on blockchain and cryptocurrency rails, where community members become competition sponsors and prize pools are paid in BGR tokens. The winning models are placed on the AI marketplace for people to purchase API calls to utilize the AI models, which provide Competition Sponsors and Data Scientists a source of revenue. Refer to the tokenomics section for more details.

Data Science Competitions Life Cycle

Our data science competitions are at the heart of our mission to democratize AI. These competitions are designed to bridge the gap between companies seeking innovative solutions to data problems and the talented data scientists eager to tackle these challenges. This section outlines the structured process of how data science competitions work within the bitgrit ecosystem.

Problem Definition and Preparation

Our journey begins by collaborating with small businesses, enterprises and community members looking to harness the power of data science. We assist them in defining a problem that they need to solve or refining an existing challenge. This initial step is crucial, as it sets the stage for the entire competition. We work closely with these partners to ensure the problem is well-framed, relevant, and feasible for our community of data scientists.
For small business and enterprise competitions, the bitgrit team works with the entity to establish the problem statement and data set. For community competitions, bitgrit will use measures such as community polls or community suggestions to establish problem statements and data sets will be provided by community members, taken from open source / public data sets or acquired for a fee.
Thereafter, bitgrit takes on the responsibility of preparing the necessary data. We gather, clean, and structure the data, preparing it for analysis. This ensures a level playing field for all participants and saves valuable time that would otherwise be spent on data preprocessing.

Formulating Mathematical and Machine Learning Problems

With the problem and data in hand, a mathematical and machine learning problem is formulated. This step is essential to provide participants with a clear understanding of the task at hand. A balance between complexity and feasibility is sought, enabling both beginners and experts to participate.

Hosting the Competition

The competition phase typically spans more or less two months, during which participants, including data scientists and machine learning enthusiasts, compete to develop the most effective solutions. We offer continuous support, maintaining an open line of communication with all participants. Effective marketing efforts are deployed to engage our community of experts, ensuring a diverse range of perspectives and ideas.

Competition Judgment and Reward

Upon the conclusion of the competition, Bitgrit ensures that the top contenders receive their well-deserved recognition and rewards. We meticulously validate that the algorithms submitted comply with all the rules and guidelines. Bitgrit’s quality team of in-house Data Scientists review each model submitted in depth and score them according to the scoring sheet provided to the participating Data Scientists.
Each competition culminates with a leaderboard, showcasing the performance of all submissions. The top-performing individuals or teams (usually top 3) are crowned the winners, with the prize contingent upon:
  • Delivery of winning solutions accompanied by comprehensive documentation; and
  • Granting a global, perpetual, non-exclusive license for the client's commercial use of the solution.
The top three data scientists or competition participants are awarded an equal share from the net prize pool, fostering a competitive yet collaborative spirit within the community.
For small business and enterprise competitions, these entities provide the prize money in USD. For community competitions, community members who become competition sponsors, provide the prize pools in BGR.
bitgrit charges a commission on each competition (% on a case by case basis) as a fee for hosting and facilitating the competition.

Revenue Sharing and Marketplace Integration (Only for Community Competitions)

An exciting feature of the Community Competitions is that the winning models and algorithms are not exclusively used by one party and the rewards don’t stop at the prize pool. Instead, these winning AI models will be integrated into our AI marketplace.
This marketplace acts as a hub for algorithm creators to showcase their work, while small businesses, enterprises, and community members can access these powerful solutions.
Revenue sharing becomes a reality, as the competition sponsors and algorithm creators (Data Scientists) share the proceeds generated by the adoption of these solutions when AI marketplace customers purchase API call credits and use them on these AI models. Further details are explained under the Tokenomics section.

Data science competition has the following benefits which directly impacts on business ROI:

1. Reduction in personnel costs for machine learning model development:
Typically, hiring dedicated machine learning engineers is necessary for developing machine learning models. In recent years, there has been an increase in the application of machine learning within companies to improve services and streamline operations. Additionally, the commercialization of data science has expanded. Acquiring and retaining machine learning engineers involves significant costs. Furthermore, models built by these engineers may carry the risk of not being the best possible models developed at that time, as they are influenced by the resources available to the engineer and their own experience. Conversely, if data collection and preprocessing for building machine learning models have already been completed (bitgrit provides consulting services from data selection for competitions to data collection and preprocessing), organizing a competition immediately can allow for the development of the best model from a large number of participants compared to development by in-house dedicated engineers. This enables a significant reduction in the costs associated with hiring machine learning engineers and managing resources, while also obtaining the best model.
Additionally, conducting interviews with developers of models adopted through competitions as a secondary effect allows for gaining insights into the process of model development and acquiring knowledge for future in-house model development.
2. Recruiting machine learning engineers:
Data science competitions provide a platform to connect with and potentially hire engineers who rank highly on the competition leaderboard. Being ranked among the top performers in a competition with numerous participants indicates a high level of quality and makes it easier to screen for qualities required in potential hires, such as their areas of interest and technical knowledge.
Hackathons, which often involve engineers competing against each other to solve specific problems, are commonly held for the purpose of recruiting engineers who excel in certain domains. In this sense, data science competitions can also be seen as a platform for gathering engineers well-versed in specific fields of machine learning. Additionally, many competition participants include competition achievements on their resumes, which can facilitate the hiring process as there is alignment of interests between the sponsoring company and the winners.
Competition Topics
There are two ways in which bitgrit prepares challenges for data science competitions. One is when a sponsor company already has a problem they want to solve. The other is when bitgrit independently sets challenges that are either currently needed or likely to be needed in society.
When a sponsor company already has a problem they want to solve, the process is straightforward. The challenge is used in the competition as it is, or with modifications (if the company does not want to publicly disclose their own challenges), according to the needs of the sponsor. On the other hand, when bitgrit independently sets challenges, they are based on discussions among members with diverse backgrounds such as the marketing department and machine learning engineers. The goal is to set challenges that are likely to become popular topics or will be needed in society in the future, and the data is prepared through unique channels.
The steps to join the competition
The steps for a data scientist to participate in a competition are fairly straightforward. First, they register their email address on bitgrit's data science platform. Once the registration is complete, they can access the competition data by accepting the NDA for the ongoing competition and participate. Please note that it is not possible to register multiple accounts using the same email address, and participating in competitions with multiple accounts using multiple email addresses is strictly prohibited.
Leaderboard
The "leaderboard" in a data science competition is a ranking system that publicly displays the performance of participants' models based on specified evaluation metrics such as accuracy or F1 score. The leaderboard indicates how well participants' models perform compared to other participants. The benefits of using a leaderboard in a competition are as follows:
1. Motivation: The leaderboard stimulates competitiveness and provides participants with the motivation to strive for better results. Seeing their own rank and its changes on the leaderboard serves as a driving force for participants to invest time and effort into improving their position.
2. Visibility: High ranks on the leaderboard enhance visibility and recognition within the data science community. Participants' names may gain attention and they may be recognized as exceptional data scientists by companies and potential employers, contributing to the activation of both participants and the community as a whole.
3. Eligibility for rewards: In bitgrit competitions, participants who achieve top rankings are eligible for rewards. The leaderboard plays a critical role in determining the winners (as well as detecting fraudulent activities) and deciding the eligibility for rewards. This aspect is important for participants to compete either for recognition or monetary rewards.
As a new feature to further enhance data science competitions in the future, it is also being considered to provide a platform for participants to exchange information. In such a platform, participants who have achieved top rankings can engage in discussions with one another, allowing them to learn approaches and techniques they may not have been aware of. Such an exchange platform provides a cooperative environment where participants can discuss strategies, share ideas, and learn from each other.