Agentic AI Hackathon

Agentic AI Hackathon

Tech Zephyr 4.0 | IIT Bhubaneswar

Build AI Systems That Act, Adapt, and Execute.

Most AI applications stop when the model generates an answer.This hackathon asks you to build systems that don't. The Agentic AI Hackathon challenges participants to develop AI systems capable of pursuing a goal through multi-step decision-making, tool interction, adaptation, and autonomous execution.

Join the Whatsapp Group: https://chat.whatsapp.com/ICbOuGu5tLB3BiKVtZO0oi?s=cl&p=a&ilr=1

Build an intelligent system that can observe → decide → act → evaluate → adapt.

The problem is yours to choose. The architecture is yours to design. The solution must prove that agentic behavior is actually necessary for the problem you are solving.

The Challenge:

For this hackathon, an agentic system should demonstrate the following characteristics:

• Goal-Driven Execution
• Dynamic Action Selection
• Multi-Step Execution
• Adaptation
• Robustness

Note: Multi-agent architectures, vector databases, RAG pipelines, long-term memory, or specific frameworks are not mandatory.

Problem Domains:

Participants may choose their own problem domain.

Possible areas include, but are not limited to:

• Education
• Finance
• Healthcare
• Research
• Business Operations
• Cybersecurity
• Scientific Computing

Participants are encouraged to identify high-friction workflows where autonomous decision-making can provide meaningful value.

Eligibility & Team Structure:

• Open to all AI enthusiasts.
• Students from different institutions may form a team.
• Inter-disciplinary teams are encouraged.
• Team Size: 1–4 members.
• Each participant can be a member of only one team.

Technology Policy:

Participants are free to choose their technology stack.

You may use:

• LLM APIs.
• Agent frameworks.
• RAG systems.
• Vector databases.
• External APIs.
• Open-source libraries.
• Publicly available datasets.
• Cloud services.
• AI coding assistants.

Examples of permitted tools include LLM APIs, agent frameworks, databases, search APIs, and code-execution environments.

Stage 1 — Online Submission:

Teams must submit the following:

Problem & Solution Brief:

A concise document or presentation containing:

• Problem statement.
• Target users.
• Why the problem requires an agentic solution.
• Proposed solution.
• Expected impact.

System Architecture:

Provide an architecture diagram showing the major components of your system.

The diagram should explain, where applicable:

• Agent/controller
• Tools
• External systems
• Memory/state
• Retrieval
• Planning
• Evaluation/verification
• Human interaction
• Failure handling

Source Code:

Submit a public GitHub repository containing:

• Source code
• Setup instructions.
• Dependencies.
• Environment configuration instructions.
• README.
• Architecture documentation.

Demo Video:

• Submit a 3–5 minute screen recording demonstrating one complete workflow.
• The demonstration should show:
• Goal → Decision → Action → Intermediate Result → Adaptation → Final Outcome
• Teams are encouraged to demonstrate at least one failure or unexpected condition and show how their system responds.

Deployment / Runnable Version:

Teams should provide a way for judges to run or interact with the system.

This may be:

• Web application
• API
• Hosted application
• Runnable notebook
• Local setup
• Other reproducible implementation

If deployment is technically impractical, teams must provide clear instructions for reproducing the demonstration.

Stage 2 — Grand Finale

The Live Challenge:

At the beginning of Stage 2, a brand-new problem statement will be released.

The challenge will strictly revolve around designing, building, or adapting an Agentic AI Chatbot to solve the newly presented problem.

Teams will pitch their newly built Agentic Chatbot to the judging panel.

Format: 7-Minute Presentation & Live Demonstration + 5-Minute Technical Q&A

Key Offline Requirements for Finalists:

Hardware & Environment: Teams must bring their own hardware setup and API credentials.

Live System Walkthrough: Finalist teams must run a live, unscripted query through their chatbot in front of the jury. Mocked responses or pre-recorded videos will not be accepted for the Round 2 challenge.

Code Of Conduct:

• All participants must maintain a respectful and inclusive environment.
• Harassment, discrimination, or inappropriate behavior will not be tolerated and may result in disqualification.

General Terms:

• Organizers reserve the right to modify the rules, disqualify participants, or cancel the event in case of unforeseen circumstances.
• All decisions by the judging panel and organizers are final and binding.

Prize pool: Worth 35,000 (INR)

Instagram: https://www.instagram.com/techzephyr.iitbbs/