BotUniversity

Accelerate your AI development by outsourcing the data labeling process

Fast

We multiply the pace of labeling processes with strong automation and the help of German Students.

Secure

Our Students create your labeled dataset in closed and secure environments we call BotUniversities.

High-Quality

To ensure the highest possible quality, we pick students from task matching courses and add extensive training.

Problem

Data annotation accounts for about 60-80% of the cost and time of AI-projects (100k to 10M annotated samples are needed for a well-performing AI), while being a highly repetitive, comparitively boring and sometimes undemanding task. Available project staff (like data scientists) is overqualified and overpaid to do this work. Outsourcing to crowdsourcing platforms, such as Amazon Mechanical Turk, is cheaper but results are often of questionable quality and the data could get indirectly publically published.

The BMWi Monitoring Report Wirtschaft Digital has identified data security concerns and cost as the main hindering factors that keep German enterprises from implementing AI, so what is missing is a solution that combines the high speed and low cost of crowdsourcing with the security and quality of in-house annotation. We, the BotUniversity, plan to fill this gap, while helping to lead the AI-project from idea to implementation.

A quick example:

A medium-sized automotive supplier wants to use AI to automatically check welds on workpieces for detecting defects.
Since only low error rates are tolerated in the automotive industry, the AI has to work extremely accurate, therefore a data set of about 1 million samples is required. A team of 10 engineers (monthly salary 5500€ gross for employers) could process this dataset (1 minute per label, 8h/day, 20 days/month) in 10.5 months. This would cost the company 577.500€ in salaries alone. Due to these costs and the lacking availability of the engineers for other tasks it is currently not attractive for the company to implement such an AI.

1 Million

Samples

1

Label per min

16.670

Hours Of Labour

577.500

Euros

Solution

With the help of students from German universities, our strong active-learning system and a clever software interface, we are able to annotate data at least 5x faster and 25% cheaper compared to inhouse labeling, while we focus on impeccable data security and expert training for our students.

Sequel to the example:

The BotUniversity will be a closed room close to a German University (starting with the University of Bremen), where 30 students label Your data in parallel. The room with access restriction will be open everyday for registered students. We value expert training, reward great performance and ensure a good working environment for our students, so that they can deliver great labels. For this specific example, we would pick students from engineering courses who already have prior knowledge on how a good weld would look like. These prospective experts deliver great annotations with high consistency. By high parallelization of the task, the aid of our active-learning system and a well-optimized software interface, we can reduce the time needed for the specific task from 10.5 months to 1.5 months while reducing cost.

10.5

Months

1.5

Months

Four steps towards your labeled dataset:

① We collect your requirements and data.

② We brief the students on the requirements.

③ Our students create the labeled dataset.

④ We verify the quality and return the dataset.

Our Team

Jannes Nazarek

Co-Founder - CEO

Maik Daniel Klause

Co-Founder - CTO

Felix Müller

Co-Founder - CPO

Get in touch

Phone

Mo - Fr 8:00-18:00

Email

contact@botuniversity.io

Address

BotUniversity UG (haftungsbeschränkt)
Martinistrasse 62-66, 28195 Bremen, DE

* Required fields

We make Artificial Intelligence smarter