AI Enterprise SolutionsGPT & Claude

Harmony

Teams often discover project delays only after deadlines have been missed. Traditional task management tools help track assignments, deadlines, and progress, but teams still need to manually identify warning signs, understand dependencies, and figure out what might go wrong next. Foresight is an AI-powered proactive project risk management platform that analyzes task deadlines, progress, priorities, recorded activity, and dependencies to identify potential problems early. It assesses how one delayed task could affect other tasks or project milestones, explains the evidence behind each warning, and recommends practical preventive actions. Instead of simply reporting what is overdue, Foresight helps teams understand what may go wrong, why it matters, and what they can do about it. Personalized insights are delivered through an AI Inbox, where team members can review, acknowledge, and act on recommendations. Our goal is to help teams move from reactive task tracking to proactive, data-informed project management.

Open the demo Video GitHub

Preliminary score

34.5/ 100Earlyaverage of 2 judges

Value for the user13.5 / 25
A specific problem and a clear outcome from solving it.
Prototype and use of AI6.5 / 30
A working core scenario, and what AI actually contributes to the result.
Quality testing5.0 / 20
Tests, examples of failures, and comparison with the current approach.
Feasibility6.0 / 15
Data requirements, running costs and a clear next step.
Originality3.5 / 10
A meaningful difference from common solutions.

What the judges wrote

GPT-6 AstraPreliminary33/100
Strengths
Your proposal addresses a real operational problem: discovering delays too late to prevent downstream impacts. Separating deterministic risk rules from evidence-grounded explanations is a sensible architecture. Your disclosure of the Member.getId() compilation issue and the absence of measured prediction accuracy is appropriately transparent.
To improve
Show a reproducible end-to-end execution from task records through dependency analysis to a persisted inbox warning, confirm the compilation fix, and replace the model placeholders with the actual runtime integration status. Turn your planned checks into executed tests with expected and actual results, including completed-task exclusions, dependency propagation, and duplicate notifications, and compare against ordinary overdue-task tracking. Name a specific adopting role and measurable adoption outcome, and document exact setup requirements and estimated operating costs.
Claude Opus 5.5Preliminary36/100
Strengths
You frame a real pain clearly: teams find out about delays only after deadlines slip. Detecting risk with deterministic Java rules and keeping the LLM for evidence-grounded explanations is a sensible, auditable design choice for enterprise use. Tracing how one delayed task affects dependent tasks and milestones, and delivering it through an AI Inbox, gives the product a concrete shape. You were honest about not yet having measured prediction accuracy and about the compile-time Member.getId() mismatch you hit.
To improve
Your submission still contains unfilled template placeholders for the runtime model and the development-time AI tools. Name the exact LLM you integrated, or say plainly that none is integrated yet, because right now AI's contribution to the result cannot be confirmed. Your testing section is mostly a plan. Run the listed cases on a seeded demo project (overdue detection, dependency propagation, duplicate-notification prevention) and report pass/fail results with examples. Add the metric your track asks for, such as days of early warning compared with the current manual status review, plus rough running costs per project per month.

The team

Rubabe Qubadova
developer
Aydan İsmayilovaLead
developer
Huseyn Qasimov
developer
Mehdi Sədrəddinov
product

Disclosure

Required by the rules. Judges verify it against the repository.

Models: Foresight's core risk detection approach is based on deterministic Java business rules rather than relying entirely on an LLM to decide whether a project is at risk. The risk engine evaluates structured project data and identifies contributing factors. Where an LLM integration is enabled, it is intended to generate natural-language explanations and actionable recommendations grounded in the supplied evidence. No custom AI model has been trained or fine-tuned for this project. Runtime model/provider: [Name the exact model and provider integrated into the submitted version, or state that no runtime LLM has been integrated yet.] Development-time AI tools and models, if used: [Disclose the models actually used during development, separately from runtime models.]
Data: The platform's risk analysis is designed to use project-specific information, including task status, deadlines, assignments, priorities, recorded progress, and task dependencies. Any demonstration scenarios and test records are used to illustrate project risks and validate system behavior. No external customer dataset is claimed as part of this submission. If an external AI provider is configured, only the project information required for a particular analysis should be sent to that provider. The submitted implementation's actual data sources and AI data flow should be verified against the code and configuration.
Components: The backend is being developed in Java using the Spring Boot ecosystem, with REST APIs and a modular backend architecture. The intended stack includes Spring Data JPA/Hibernate for persistence, PostgreSQL for relational data storage, and Maven for dependency management and builds. The application architecture separates project and task management, deterministic risk analysis, AI-generated explanations and recommendations, and the persistent AI Inbox. Third-party libraries and frameworks are used according to their licenses. Any additional dependencies, public templates, AI integrations, or external services included in the submitted version are disclosed in the repository's dependency files and documentation. No previously developed product is being submitted as a new hackathon project.
Built during the event: Confirmed

Quality testing

We are validating the backend incrementally, with particular attention to deterministic risk detection and the correctness of project and task data. During early compilation, we identified a mismatch between the risk-analysis implementation and the `Member` domain model: the analyzer attempted to call `Member.getId()`, but that method was not available in the current implementation. This exposed the need to align domain models, entity accessors, and risk-analysis logic before completing verification. The next validation steps focus on: * Detecting overdue tasks while excluding completed or cancelled tasks. * Identifying approaching deadlines using available progress and activity evidence. * Tracing risks through task dependencies. * Verifying risk scores against documented rules. * Preventing duplicate active inbox notifications during repeated analysis. * Checking database persistence, authentication, and project-level data isolation. We do not yet claim measured prediction accuracy or quantified reductions in project delays. The priority is to establish reproducible risk detection, verify the backend through automated tests, and document the observed failures and fixes.

Worth 20 of the 100 points.

Running it

### Prerequisites * A compatible Java Development Kit (JDK) * Maven * PostgreSQL * Git ### Setup 1. Clone the public repository. 2. Navigate to the backend directory. 3. Configure the PostgreSQL connection and required environment variables according to the project's configuration example. 4. Initialize the database using the project's documented migration process. 5. Run the automated tests: `mvn clean test` 6. Start the backend: `mvn spring-boot:run` 7. Access the API documentation at the configured OpenAPI/Swagger endpoint, if enabled. ### Configuration Never commit real database passwords, JWT secrets, or AI API keys. Use environment variables or a local environment configuration file. The exact Java version, database settings, required environment variables, and API documentation URL are specified in the repository. Note: External AI services are optional only if the submitted implementation includes a working deterministic fallback. Follow the repository's actual setup instructions.

Project ownership

The team keeps the rights to this project. Judges were given access for verification only.