Multi-leg delivery ETA system
Inspired by publicly reported work at Swiggy
Split a delivery into kitchen, pickup and drop legs, predict each one, and keep the estimate honest as the order moves.
The brief
Swiggy has described predicting delivery time leg by leg (assigning a rider, reaching the restaurant, waiting, and the last mile) and updating the estimate with live signals, and explaining predictions with SHAP. Build your own multi-leg ETA system: model each leg, combine them, update the estimate as events arrive, and monitor it. Good looks like: each leg evaluated separately, a combined ETA that beats a single model, prediction intervals with honest coverage, explanations, and drift monitoring.
Read the case first
What Swiggy has said in public about this kind of work, on its own pages:
- Predicting Food Delivery Time at Cart (Swiggy Bytes) ↗
- How ML Powers "When is my order coming?" Part I (Swiggy Bytes) ↗
Make it your own
Apply the legs to another trip: an ambulance call, a courier pickup or a college bus. Name the legs for that trip, say which signal you would need for each, and simulate the ones the open data lacks.
AITeky has no link with Swiggy, which has not reviewed or endorsed this brief. The brief only draws on what the company has published, uses open data rather than its data, and is a learning exercise: don't present your project as Swiggy's work or system.
What you will build
- Define the legs and their targets from the data
- Train a model per leg and a single-model baseline
- Combine legs and add prediction intervals
- Update the ETA as order events arrive (simulated)
- Explain estimates with SHAP
- Serve the ETA through an API with monitoring and drift alerts
Free, open data to build it with
Sign up free to get the data and links
It takes a minute with your email. AiTeky Pass (₹999/month or ₹9,999/year) is free for your first 30 days.