KERNEL '26MACHINE LEARNING
Data Nilaam
Build the bank, buy the data
An elite machine learning challenge where teams run NovoCredit, a simulated digital lender, and bid on masked dataset features in a live auction. Participants must balance feature acquisition budgets against predictive accuracy to maximize credit card portfolio profit on hidden test sets.
EVENT DETAILS
📅Prelims: Data Auction: TBD
📅Finals: Model Submission: TBA
🕒10:00 am - 12:00 pm (Auction) · 12:30 pm - 4:30 pm (Build)
📍Smart Lab, Dept. of IT, Salt Lake Campus
🌐Hybrid
👥2-4
PRIZE POOL
₹6,000
EVENT RULES
- •Teams must consist of 2 to 4 members.
- •Each team is allocated a fixed bank of virtual auction tokens to bid on masked feature blocks.
- •No team may exceed their virtual budget; strategic bid allocation is required.
- •Final models will be evaluated against an unseen private ground-truth test dataset.
- •Teams must submit clean code notebooks and a short auction strategy report.
EVENT FORMAT
Round 1: Live Interactive Data Auction
- •A real-time auction room where feature subsets, correlation matrices, and noisy signals are auctioned off in 15-minute intervals.
- •Teams must decide which data to buy and which to skip under time pressure.
Round 2: ML Pipeline & Risk Evaluation Benchmark
- •Teams engineer features and train predictive credit-scoring models with their purchased dataset.
- •Submissions scored on ROC-AUC, profit simulation, and default risk penalty.
CONTACT
Rishav Chowdhury📞8276991374
Debarun Patra📞8582922525
