TRACK 01
Data Intelligence Agent
Turn complex enterprise data into actionable insights.
REGISTRATIONS CLOSED
72-HOUR BUILD SPRINT / 02–04 OCTOBER 2026
Build the next generation of AI agents.
Turn data into intelligent systems that can reason, retrieve, act, and learn.
02–04 OCTOBER 202672 HOURSAI × DATA × AGENTS
Registration for the Databricks 72-Hour AI Agents Hackathon has now closed.
TOP 50 TEAMS→PPI OPPORTUNITY
AGENT_EXECUTION / ARCHITECTURE● SYSTEM ONLINE
USERintent / context
AI AGENTplan / reason
RETRIEVALgrounded context
DATABRICKSgoverned data
TOOLSexternal actions
EVALUATIONquality / safety
ACTIONmeasurable outcome
DATA → INTELLIGENCE → ACTION72H / BUILD
EVENT COUNTDOWNUntil the 72-hour build sprint begins
00DAYS
:00HOURS
:00MINUTES
:00SECONDS
72HOURS
50PPI OPPORTUNITIES
2–4TEAM MEMBERS
1BUILD SPRINT
01 / THE CHALLENGE
Don't build another chatbot.
Build an agent that can reason, retrieve, act, and prove its work. Start with a real business problem; connect data, tools, and evaluation into a working system.
RETRIEVALREASONINGTOOL USEORCHESTRATIONSTRUCTURED OUTPUTSEVALUATIONOBSERVABILITY
TRACK 02
Research Agent
Research, synthesize, validate, and cite information.
TRACK 03
Operations Agent
Identify operational problems and recommend or execute actions.
TRACK 04
Analytics Copilot
Make complex datasets accessible through natural language.
TRACK 05
Open Track
Have a different idea? Build it. Make the architecture and impact clear.
02 / TECHNOLOGY
Build on the Databricks
Data + AI Platform.
Use the Databricks platform to move from raw data to production-minded AI systems.
REFERENCE ARCHITECTURE
From signal
to decision.
Start with reliable data. Every useful agent needs context it can trust.
SELECTED STAGEDATA
03 / SPRINT TIMELINE
72 hours. Three phases.
From first architecture to final demonstration, the sprint is designed around visible engineering progress.
PHASE 01
02 OCTOBER
Discover
- Kickoff
- Challenge reveal
- Team formation
- Architecture planning
- Data exploration
PHASE 02
02–03 OCTOBER
Build
- Prototype
- Integrate data
- Build agent workflows
- Evaluate
- Iterate
PHASE 03
04 OCTOBER
Ship
- Final testing
- Submission
- Demo
- Judging
- Top 50 announcement
04 / BUILD CHECKPOINTS
More than a GitHub repository.
Builds will be reviewed throughout the sprint, not just at the finish line. Every checkpoint captures the team's thinking and the state of the system.
CHECKPOINT 01PLAN
Architecture Lock
- System architecture
- Agent workflow
- Data flow
- Team responsibility map
- Physical architecture-board snapshot
- 2-minute architecture walkthrough
CHECKPOINT 02BUILD
Live Build Review
- Working prototype
- Current agent execution
- Progress recording
- Debugging notes
- Updated architecture
- Team walkthrough
CHECKPOINT 03SHIP
Ship Review
- Final working system
- Final architecture
- Evaluation results
- Failure cases
- Iteration summary
- Final team demo
PLAN→BUILD→REVIEW→ITERATE→SHIP
05 / COLLABORATIVE BUILD
Build together.
Ship together.
Maintain a live build record. These artifacts make decisions visible, sharpen handoffs, and help the whole team explain what changed.
Architecture Board
A visual map of the system and agent workflow.
Decision Log
Major technical choices and why they were made.
Iteration Record
What changed between checkpoints.
Team Demo
A short demonstration of the working system.
Role Map
What each team member contributed.
ARCHITECTDATAAGENTEVALUATIONPRODUCT
06 / FINAL SUBMISSION
Ship something real.
A working system, a clear architecture, and evidence that your team evaluated and improved it.
“Show us how your system evolved from your first architecture to the final build — and why.”
EVERY TEAM SUBMITS
- Working application
- GitHub repository
- 3-minute demo
- Architecture diagram
- Technical README
- Evaluation report
- 90-second build story
- Checkpoint artifacts
07 / JUDGING
How your build
will be evaluated.
Don't just demo an agent. Measure it. Judges will look at the system, its evidence, and the engineering decisions behind it.
Problem & Impact25%
Technical Depth25%
Agent Quality20%
Data & Evaluation15%
Product Experience10%
Build Process5%
QUESTIONS WE'LL ASK
Does the agent actually reason?
Is the data pipeline meaningful?
Are outputs measurable?
Does the system handle failure?
Is the architecture understandable?
Is Databricks used thoughtfully?
Can the prototype evolve toward production?
08 / THE OPPORTUNITY
50
Your build could be your next opportunity.
Top 50 teams receive a PPI opportunity based on their hackathon performance.
Event Status ↗PATH TO THE TOP 50
ALL TEAMS
↓
SUBMISSIONS
↓
TECHNICAL EVALUATION
↓
TOP 50 TEAMS
↓
PPI OPPORTUNITY
09 / RECOGNITION
Work worth recognizing.
PPI Opportunity
Based on hackathon performance
Technical Recognition
Technical Recognition
Technical Recognition
Community Recognition
10 / PARTICIPATION
Rules of the sprint.
Clear requirements help teams focus on engineering, collaboration, and a working demonstration.
Teams consist of 2–4 members.
The hackathon runs for 72 hours from 2–4 October.
Teams must complete all required checkpoints.
Every team must submit a working demonstration.
Every submission must include an architecture diagram.
Databricks technologies should be meaningfully incorporated.
Open-source models and third-party APIs may be used where permitted.
AI coding assistants may be used, but participants must understand their implementation.
Document external datasets, models, APIs, and dependencies.
Teams must explain their architecture and technical decisions.
Substantially copied or plagiarized submissions may be disqualified.
PPI opportunities are based on hackathon evaluation.
11 / FAQ
Good questions.
Clear answers.
EVENT STATUS / REGISTRATIONS CLOSED
Registrations
are closed.
Registration for the 2–4 October hackathon has ended. Registered teams can access their participant dashboard for challenge details, build checkpoints, and submission requirements.
EVENT STATUSREGISTRATIONS CLOSED
HACKATHON02–04 OCTOBER 2026
FORMAT72-HOUR BUILD SPRINT
PPITOP 50 TEAMS