Build AI That Thinks. Acts. Verifies.
Enter the unknown. PROJECT: ECLIPSE is a technical hackathon challenging developers to construct autonomous agentic systems that execute real workflows, use APIs, verify outcomes, and handle edge cases.
How True AI Agents Work.
PROJECT: ECLIPSE is designed around agentic execution. Teams must build systems that follow a complete loop from initial input to verified real-world recovery.
INPUT
Ingest raw data, alerts, or legal contracts
UNDERSTAND
Extract context, entities & constraints
PLAN
Formulate execution strategy & hypotheses
USE TOOLS
Call APIs, databases & system tools
ACT
Execute autonomous operational tasks
VERIFY
Confirm state changes & system outcomes
ESCALATE
Human-in-the-loop approval for high impact
Beyond Generic Chatbots
Judges look for autonomous execution, tool integration, and failure recovery — not static prompt wrappers.
The Challenges.
AI Agent for Production Incident Investigation & Recovery
Autonomously investigate production incidents across multiple data sources and orchestrate recovery.
Autonomous Resolution of Cross-System Transaction Mismatches
Match and resolve transaction discrepancies across financial systems with an auditable AI agent.
From Contract Language to Executable Business Actions
Extract obligations from contracts and automatically generate actionable business workflows.
Autonomous Requirement-to-Quotation Procurement System
Transform procurement requests into evaluated quotations with an autonomous AI agent.
Issue-to-Pull-Request AI Engineering Agent
Autonomously resolve GitHub issues and generate review-ready pull requests.
OPEN-X — Open Innovation — Build Beyond the Brief
Define your own real-world problem and build a GenAI, Agentic AI or intelligent automation solution.
Hackathon Timeline & Rounds.
Track key milestones across each round from registration to the grand finals.
Registration & Idea Submission
Register your team (1-5 members) on Unstop or HACK2SKILL. Choose your challenge track or OPEN-X.
Agentic Core Build
Construct initial autonomous workflow, tool integrations, and basic verification loop.
Scaling & Edge-Case Handling
Handle edge cases, multi-step failure recovery, human-in-the-loop escalation, and live trace logs.
Live Demos & Award Ceremony
Finalist teams present live agentic execution to the judge panel for final scoring.
How Projects Are Evaluated.
All submissions are scored across 7 weighted dimensions totaling 100%. Judges prioritize genuine tool execution, failure handling, and multi-step reasoning.
- 01Problem Understanding & Relevance10%
Does the team clearly articulate the problem domain and why their solution addresses it?
- 02Agentic Reasoning & Planning20%
Does the system demonstrate autonomous multi-step reasoning, decomposition and planning — not just prompt-response?
- 03Tool / API / System Integration15%
Does the agent call real tools, APIs or systems and act on the results?
- 04Autonomous Task Completion20%
Does the system complete meaningful end-to-end tasks without constant human prompting?
- 05Reliability & Handling Failure Cases15%
How well does the system handle unexpected inputs, API failures, ambiguous data and edge cases?
- 06Innovation10%
Does the solution introduce a novel approach, architecture or application of AI/agentic systems?
- 07UX / Demonstration10%
Is the demonstration clear, compelling and does it show the system working end-to-end?
| No. | Judging Criterion | Weight | Evaluation Description |
|---|---|---|---|
| 01 | Problem Understanding & Relevance | 10% | Does the team clearly articulate the problem domain and why their solution addresses it? |
| 02 | Agentic Reasoning & Planning | 20% | Does the system demonstrate autonomous multi-step reasoning, decomposition and planning — not just prompt-response? |
| 03 | Tool / API / System Integration | 15% | Does the agent call real tools, APIs or systems and act on the results? |
| 04 | Autonomous Task Completion | 20% | Does the system complete meaningful end-to-end tasks without constant human prompting? |
| 05 | Reliability & Handling Failure Cases | 15% | How well does the system handle unexpected inputs, API failures, ambiguous data and edge cases? |
| 06 | Innovation | 10% | Does the solution introduce a novel approach, architecture or application of AI/agentic systems? |
| 07 | UX / Demonstration | 10% | Is the demonstration clear, compelling and does it show the system working end-to-end? |
Compete & Win Big.
Recognizing top agentic AI innovations. Claim your spot on the podium and win cash prizes, certificates, and mentorship.
Built For Serious Engineers.
This is not a wrapper hackathon. We evaluate real architecture, autonomous decision loops, system integration, and fault tolerance.
REAL-WORLD COMPLEXITY
Solve actual enterprise bottlenecks across operations, finance, legal, procurement, and software engineering.
TOOL & API CALLING
Integrate external services, databases, webhooks, and REST endpoints into autonomous agent workflows.
VERIFICATION & RECOVERY
Build systems capable of verifying their own actions, analyzing execution failures, and iterating to self-heal.
GOVERNANCE & ESCALATION
Implement safety guardrails and human-in-the-loop approval workflows for high-impact operational decisions.
Don’t just show what your AI says.
Show what your AI does.
Ready to enter the unknown?
Registration is now open. Form your team, select your challenge track, and build next-generation agentic AI systems.
Everything You Need to Know.
Got questions about PROJECT: ECLIPSE? Find clear answers regarding tracks, registration platforms, guidelines, and evaluation.