Overview

Timeworx.io is a powerful, scalable, and future-proof platform for businesses to process data, either specific datasets or real-time data, easily and accurately by using a decentralised workforce of agents incentivised via Blockchain technologies. We unleash the power of human intelligence to provide performance validation, and training & fine-tuning for Machine Learning and AI systems to ensure accuracy and relevance.

The core of the platform is the decentralised protocol for data processing that is governed by the TIX token for ensuring transparency, traceability and fairness. Any business, acting as a customer of the platform, can request data processing services using TIX, and can submit all of its collected datasets along with instructions on how the data is intended to be processed. From then on, the protocol binds Agents as service providers that process the data as intended in exchange for TIX.

The platform defines the following protocol for data processing, as depicted in Figure 4:

1. A business submits a dataset in the platform and specifies how the data needs to be processed.

2. The platform then goes through the entire dataset, taking each data item at a time and generating a data processing Task for it. You can think of a Task as a set of specific instructions that need to be performed on a specific piece of data, e.g. “What emotion do you feel when you read the following text? Positive, Neutral or Negative?”

3. All of the generated Tasks are then distributed to a scalable pool of Agents

4. Each Agent picks up a Task, processes the data according to the instructions and proposed a Task Output in the the protocol

5. Finally, the protocol aggregates all of the Task Outputs and delivers the processed results back to the business.

Fig. 4. The Timeworx.io data processing protocol

The decentralised protocol is designed to ensure that businesses obtain the ground truth from data processing through a trustless implementation: no Agent is considered trusted. Therefore, multiple copies of the same Task (associated with the same piece of data) are distributed to multiple Agents at the same time, while ensuring that the same Agent is never served with the same Task twice. After all Agents have finished processing a given Task and have proposed an output, the result of the data processing is established by running a Consensus algorithm on all of the outputs.

Based on the result obtained from the Consensus, all of the outputs are marked as correct, if they match with the agreed upon result, or incorrect otherwise. Aside from ensuring proper data processing, the protocol uses this workflow to also assess the performance and accuracy of all Agents, enforcing accountability across the platform.

Agents in the platform can be either human beings that process data manually using the Timeworx.io mobile app, or AI models that have been integrated into the protocol for processing the data automatically. Based on the type of Agents that are handling the data processing, the platform defines two modes of operation:

1. Batch (or Deferred): the data is collected and processed into batches, or datasets. This type of processing is generally used when the data is processed by Human Agents, since distributing all of the Tasks, waiting for them to be processed, running the consensus and gathering the results are time-consuming operations.

2. Real-time (or Continuous): the data is streamed continuously into the platform and processed within seconds. This type of processing is generally used when the data is processed by AI Agents which are able to perform operations at high speeds, with no overhead.

To make it easier to wrap our heads around this, let’s take a look at an example.