AI Engineering

What are AI engineering and agentic AI?

AI engineering is the practice of designing and building software systems with technologies such as large language models, AI agents, APIs, tool calling, information-retrieval systems and vector databases.

An AI agent does more than generate a response to a question. It can use tools, retrieve information, interact with external systems and perform a sequence of actions to complete a task. Technologies such as the Model Context Protocol, or MCP, can be used to connect agents to additional tools, data sources and services.

In this track, the agentic system should form the core technical challenge of the application. The project should go beyond a standard chatbot and require students to make substantial technical decisions about agent behaviour, system architecture, data retrieval, tool integration and reliability.

About the track

Each team will have five weeks to build an MVP based on an idea it has proposed. The application must use modern, LLM-based AI systems, with AI agents performing an essential part of the work. Teams are expected to design and implement the complete technical solution, including the application architecture, agent workflows, tool integrations, data storage and retrieval, and the interaction between the AI system and the rest of the application.

Possible technical components include:

• one or more AI agents;
• tool calling and external API integrations;
• MCP servers or clients;
• vector databases and retrieval-augmented generation;
• structured outputs and automated workflows;
• agent memory and state management;
• evaluation, monitoring and error handling; and
• coordination between multiple agents.

Teams are free to select the technology stack that best fits their project. Arkadia will provide a predefined number of API credits for accessing large language models.

During the track, teams will attend mentor sessions to discuss their architecture, technical choices, implementation challenges and progress. Teams should also consider the intended users and practical value of their application. During the track, they will briefly pitch the business case, receive feedback and include it in the final presentation. The main focus, however, remains the design and development of the technical system. At the end of the track, each team will demonstrate its MVP during a final presentation.

The MVP each team builds will be open sourced.

Who can apply?

The track is open to everyone who is comfortable writing Python code.

Applicants should have:

• a basic understanding of Python;
• experience working with APIs;
• a basic understanding of large language models and their use in software applications;
• familiarity with Git; and
• familiarity with team-based agile software development practices.

Applications must be submitted by a complete team.

How to apply

Submit your proposal as a PDF by 14 August 2026 to level3@arkadia.hn.

The proposal must be between one and four A4 pages and should include the following sections.

1. The proposed application: Describe the application you want to build, the problem it addresses and the intended users.
2. The agentic system: Explain what work the AI agents will perform and why an agentic approach is needed. Make clear how the application goes beyond a standard chatbot.
3. The technical approach: Briefly describe how you expect to build the system. A complete technical design is not required, but the proposal should show that the agentic system is the core technical challenge and that the team has considered how it could be built within five weeks.
4. The value of the application: Briefly explain the practical or business value of the application, why the problem is worth solving and what benefit the use of AI provides.
5. The team: Include the names of all team members and briefly describe the relevant skills or experience within the team.

Selection

Proposals will be assessed based on their originality, the value of the proposed application and the extent to which AI agents are essential to the solution. We will also consider the quality and ambition of the technical approach, and whether the project can reasonably be developed into an MVP within five weeks.

Selected teams will be invited to a short interview before the start of the track.