/ Artificial Intelligence

Part 1 - Using AI to Consistently Deliver Successful Projects and Build Teams


Taking a project to a successful market entry is a complex and costly process. Doing it alone and depending only on your personal experience and capabilities makes it even harder.

At DREAM we are combining the power of a decentralized reputation system, and A.I. to create the world’s first comprehensive, online, project accelerator ecosystem.

We are solving the problem of determining the planning and staffing requirements for complex technical projects, while only requiring semi-structured descriptions of the project’s goals, scope, and purpose.

This becomes possible by applying deep learning-based NLP (Natural Language Processing) solutions which extract and categorize information, and utilizing NLU (Natural Language Understading) frameworks to guide complex user interactions through an A.I. bot user interface.

Our matching technology is based on a hierarchical, vector representation-based recommendation system, which allows us to not only match the most suitable individual candidates to the task, but also to recommend an entire team of expert talent that has the best probability of cooperative and synergistic team dynamics.

Enabling the Community


With our solution, we are building an ecosystem of trust-enabled relationships to support project managers in getting access to the right kind of talent quickly, helping them to scale out their ideas into proven and deliverable projects.

This involves the ability to:

  • Support projects in describing and defining their vision into an actionable scope of work.
  • Enable a transparent and dependable breakdown of that broad scope to executable tasks.
  • Align the technical and non-technical requirements of the project.
  • Make use of incomplete, changing, and uncertain information to guide decision making under natural uncertainty.
  • Provide a scalable, real-time architecture and integrate external information defining the context of a project.
  • Learn from the interactions with the relevant context and improve predictive accuracy in recommendation.
  • Secure a privacy-enabled pathway to aggregate and redistribute the learning from multiple parallel projects in the ecosystem to the benefit of everyone involved.

DREAM is building a rich, orchestrated ecosystem of complementary tools, which allows any project manager (and entrepreneur) to accelerate an idea to a working product-based on the learned success of the community. This is now possible, as we stand on the vertex of two key technologies: Applied artificial intelligence and blockchain.

Project Planning and Scoping

At its core, project management is about constraining an idea to the possibilities and limitations of a specific context, resulting in a sequence of logical, executable tasks. So, the most important step at the start of every successful project is to identify the context you are working in.

DREAM Planner is an advisor, project manager, and team-building application, that enables users to quickly and iteratively move from their project idea to a successful execution.

Through a conversational interface, integrating a chatbot and a graphical user interface (UI), we enable the handling of complex, real-life project management situations in a non-linear, adaptive workflow.

DREAM Planner acts as an integrated application layer, empowering the user to deploy advanced dynamically generated analytics and Process Management Solutions as part of our Core Service Architecture, on-demand and scalable to their project needs.

DREAM provides a clean, simple interface which allows project managers to logically and sequentially feed the system information about their intended goals, moving their project from an idea, through planning, to a working product.

Knowledge about current project success factors is gleaned from the interactions of projects and distributed teams on the platform in the background, and made available through our Augmented Intelligence Process. This supports critical phases of decision making by guidance and information presentation, along proven stage gate procedures, developed with leading industry experts.

Critical learning previously lost from project-to-project is captured and used to deliver qualified answers and validated solutions to clients.

Project Data is stored in DREAM Manager, a “living business case”, that is used to keep state, storing context-specific learning from the ongoing interaction around the project in the ecosystem to continuously personalize the user experience.

Multi-threaded conversations are supported through a context-aware deep learning QA System, built on state of the art end-to-end weighted selective Attention Architecture. Model Learning is conducted on augmented datasets.

Product Architecture


  1. Display

DREAM Planner consists of two integrated input and display channels, the chatbot application and the GUI, which are consistently updated along the user interaction.

  1. Business Logic

Input to the chatbot interface is processed for intents in the NLU engine. Explicit commands, or intents calling for direct service execution without logical gate processes, are passed directly to the DREAM Builder service API for execution.

Complex intents triggering expert-trained logical decision trees are processed by the Expert System, which takes over control and manages the state of the interaction within the context variable.

Input to the GUI is processed within the logic of the display sub-services, and updates are written to the interaction context variable. This update in interaction state is promoted to the chatbot interface to enable the user to quickly switch between graphical input and written commands.

  1. Analytics Services

DREAM Builder deploys low-level services that are dynamically generated for the current state of the interaction. The low-level services execute them accordingly. Integrated responses are created to ensure the update to the conversational interface is happening in parallel.

Making it Easy to Gather Data


Creating a project is an iterative process; nobody gets it all right the first time around. So, we are building for ease of use, especially when it’s time to collect information we need to provide our personalized services.

Throughout the process of developing our solution, we will be working to go from a structured input to an ever more adaptive and responsive interaction, which will allow us to only gather exactly the information we need at a given point in time.

  1. Form-based assessment
    Data necessary to fill the slots during defined process workflows are requested from user through display of forms.
  2. Extraction-supported assessment
    Data entry to forms can be substituted with a user-supervised document extraction, and following submission to the system.
  3. Dynamic interaction-based assessment
    Data is requested by the system based on its current needs, combination of forms, and data extraction support. Manual conversational entry is supported within the workflows.
  4. Adaptive assessment
    Available data is used to predict missing data. The importance of additional information to substantiate the information estimates is calculated, and the most important clues are requested as input from user to finalize assessment.

Building for flexibility and multi-tasking

We follow a three-tier architecture with a service API for integration with the currently most relevant, professional chatbot frameworks. This enables the user to effortlessly switch between different channels most suitable for the current stage of work.

This will allow us to provide our advanced solutions to project management and scoping outside of our native web application, when our users are discussing and scoping in Slack or Telegram.

Architecture Overview


Application Layer

The use of a unified service API enables us to maximize control over the behavior of third-party chatbot frameworks, according to the given state of the conversation. This enables users to continue working in their existing workflows with the added value of the DREAM Services.

Delivery & Processing System

Our multi-threaded conversation architecture used for the NLU Engine, enables the user to follow an agile process of managing multiple objectives at the same time. Each interaction can be stopped at any time to be picked up later.

The NLU engine employs a set of domain-specific intents created by human experts and a supervised intent classification model trained on a set of high-quality interactions sequences.

A conversational state storage is employed to keep track of conversational flow, providing the basis for the routing and execution system and the needed contextual responses.

Log Store for Historical Data

DREAM stores the log data of interactions to be used for continuous quality assurance and supervised training of the NLU Engine and the deep logic network. This includes the confidence/quality metrics for the classification as well as the given responses.

Deep Logic Network Layer

The employment of logical-based expert systems, in combination with an intuitive, graphical administrative back end system, enables us to provide even highly-complex interaction sequences through a chatbot interface at scale. Domain experts model best-practice guides, which are in turn executed by a rule-based chatbot system. The resulting interactions are continuously qualified and used for the ongoing training of a probabilistic graphical model based ML system.

Machine Learning System

DREAM employs a collection of unsupervised models for training data generation and quality control. Supervised models train the intent classification and the business logic on the generated transaction data within the DREAM Ecosystem.

Business Logic Layer

It encompasses not only the business logic, but also the interface to the sub-systems necessary to fulfill the request. This involves both routing, management, and optimization of the integrated service calls.

Routing and Execution System

Given the current state of the conversation and the classified intent, the system generates a protocol to call the necessary low-level services and generates the response.

To continue our exploration of how DREAM will take project management and team-building to the next level, please click here to read part 2 of our series.

The DREAM platform isn’t just another untested beta program on a white paper… It’s live and being used right now to hire blockchain professionals. The token sale will enable DREAM’s innovative team to take DREAM to the next level by integrating A.I. and incorporating our platform token.

Social Channels

Part 1 - Using AI to Consistently Deliver Successful Projects and Build Teams
Share this

Subscribe to DREAM Blog