About the challenge
A team sprint for Cornell Systems Engineering students. Your team picks a prompt, invents the data, and ships a working quality-control tool.
- Who: Cornell students in the Systems Engineering MEng/MS program or enrolled in SYSEN 5300 / MAE 5390 or SYSEN 5900 — up to 5 per team
- Where: Upson Hall 116
- Submissions: on Devpost — the event page and submission instructions are shared at the event
-
On-Campus Version: 24 hours, October 16, 3PM - October 17, 3PM
- required for on-campus students, unless excused
-
Virtual DL Challenge: 7 days, October 16, 3 PM - October 23, 3 PM
- required for distance learning students and any others who cannot attend in person.
You will tackle one real-world quality control and reliability problem, drawn from industry, healthcare, energy, or infrastructure.
- 🧩 Prompts are released at kickoff — each tied to a dataset you design and build yourself
- 📊 The statistics are the graded core. Statistical process control, process capability, reliability modeling, failure analysis. The app is the delivery vehicle for the analysis, not the point.
- 🧱 Ship a dashboard / web app (React, Shiny, etc.) that answers the prompt, and anything you need to make it work (eg. an R or Python library, a public REST API, a design scheme, etc.)
- 🚀 Deploy it live to the course Posit Connect server — publisher credentials are handed out at the event
- 🔢 Every project is scored 0-100 by the event staff. Read the criteria
- 🏆 Top team wins a prize and bragging rights
Get Started
- 🗓️ Schedule of Events (24 hour event)
- 🧱 Slides Overview
- 💬 Prompts — revealed at the start of the event (see below!)
- 🔢 Evaluation Criteria
- 📚 Resources for Building Your Tool
- 🧰 Starter Templates
- 🤖 Agent context bundle — read this before you start building
- ⁉️ FAQ
Form a team (max 5) and register through DevPost. No team? Sign up anyway and we'll match you.
Requirements
What to Build
🤖 Build with AI — bring your own agentAI-assisted development is expected and encouraged. Use whatever you already have: Claude Code, Cursor, Copilot, Codex, Gemini CLI. The skill being tested is steering a capable assistant toward statistically correct work — which is exactly the skill this course is about.
This repo ships an agent context bundle so your assistant starts oriented!
PromptsData collection challenge: These agencies have no telemetry, no app, and no dashboards yet. Students must define exactly what proxy variables field workers or citizens need to log to make the stats work.
On-Campus Six Sigma Hackathon Prompts
Urban Systems
Bikeshare Fleets
A municipal bikeshare system is suffering from a "docks full / docks empty" crisis during rush hour. Maintenance vans are dispatched blindly. Design a system tool to track dock availability, when and where stations are unusable, and support the network's performance.
Health Systems
Emergency Response
Ambulance dispatch to on-scene times are varying wildly across city zones, but the 911 center only looks at monthly averages, missing fatal outliers, not where or why these failures occur. Design a statistical tool to help manage dispatchers, response time logging in real-time, generate analyzes, and flag zones that are in need of resolution.
Energy Systems
Smart Grid Transformer Surges
Due to sudden local adoption of heat pumps, EVs, data centers, etc., local step-down transformers are blowing out. The utility has no predictive tracking. Design a system to log load and monitor/model the ability of the system to meet transformer requirements/limits to support local neighorhoods.
Industrial Systems
Drone Delivery Fleet Maintenance
A drone startup is experiencing mid-air failures, but they only investigate after a crash. Design a tracking tool for technicians to help them log wear-and-tear before and during flight, and manage and improve system performance and/or reliability.
DL Challenge Prompts
Industrial Systems
3D Printing for Aerospace
An aerospace contractor 3D-printing titanium brackets is failing client strength tests. They don't currently track enough environmental and stress parameters against final part strength. Design a tool to monitor and flag when and whether the printing process is capable of meeting strict aerospace tolerances, and help contractors resolve these issues.
Health Care Systems
Blood Banks
A regional blood bank loses thousands of dollars of supply when older refrigeration units fail or backup generators don't start during blackouts. Design a tool to track blood bank supply integrity and flag/predict failures.
Urban Systems
Civil Liberties
A civil rights NGO suspects a government agency (local, state, federal, or otherwise) is disproportionately deploying aggressive enforcement and/or surveillance outside of the high-crime contexts they were legally approved to operate in. These might include sudden ticket blitzes, arbitrary code enforcements, encampment clearing, flock cameras, or enforcement raids. The agency claims their practices are uniform and random. Design a tool that supports concerned citizens in monitoring and tracking these outcomes and taking legal action in real time.
Prizes
On-Campus Hackathon Winner Dinner
Dinner at the Statler Dining Room for your Team.
Eligible only for on-campus students!
Distance Learning Hackathon Winner Bragging Rights
Infamy for all time
Judges
Timothy Fraser
Asst Teaching Professor, Systems Engineering, Cornell University
Thomas (Chenghao) Wu
Course Assistant / Cornell University
Tolkien Bagchi
PhD Candidate, Systems Engineering / Cornell University
Judging Criteria
-
OVERALL
Overall Summary Rating: Developing (1) to Mastery (5) -
Effective, working tool
(25 pts) does it run? -
Performs valid analysis
(25 pts) Performs valid analyses relevant to quality control in R or Python -
Scope Match
(5 pts) Scope of tool/product closely matches one of the prompts. -
Clear Use Case
(5 pts) The product should have a clear user and use case in mind. -
Clear, Minimal Requirements
(5 pts) Requires a minimal, reasonable number of inputs or requirements from the user/customer. For example, requiring the customer to know traits and values about their product that they are unlikely to be able to measure should be avoided. -
Demonstration
(10 pts) A clear demonstration of the tool, using test datasets. -
Documentation
(10 pts) Excellent, easy-to-follow documentation for the tool - how does each function work, what are the inputs, parameters, etc. -
Test Datasets
(5 pts) 2-3 accurate, working test datasets -
Dataset Explainers
(5 pts) A codebook and README for your datasets, describing what each file and variable means, and any other background information necessary for collecting this data. -
Reproducibility
(5 pts) Fully reproducible code, posted publicly to a Github Repository,
Questions? Email the hackathon manager
