Produced by: BT Desk
The moment you open the app, AI identifies your exact location using advanced location intelligence, whether you’re in an apartment or a house. "It sets expectations on ETA right from the start," says Goda Ramkumar, VP of Data Science at Swiggy, in a conversation with TOI.
AI personalises the search experience based on your previous orders and preferences. It connects you with the closest dark store or restaurant that can deliver your order the fastest.
Once your order is packed and ready, machine learning determines the fastest route to deliver it to you. "ML continually checks for delays like traffic jams or bike breakdowns," Ramkumar says, ensuring that you are alerted about any changes in ETA.
ETA predictions use deep learning models to assess factors like preparation time, rider’s travel time to the store, and then to your home. "Complex optimisations run constantly to give accurate ETAs," says Ramkumar.
AI powers inventory forecasting to ensure the right products are available in dark stores. "AI-driven forecasting is especially crucial for perishable goods to lower waste," explains Rakshit Daga, CPTO at BigBasket.
Data also informs the layout of dark stores. Frequently ordered items and commonly paired products are placed strategically for quicker packaging, allowing orders to be fulfilled in minutes.
AI-powered analytics forecast peak hours and assign manpower accordingly. "Shifts are designed so the right personnel are available at the right time," Daga explains, ensuring smooth operations in the dark stores.
Deep learning assigns riders based on proximity to both the store and the delivery address. "AI even selects the right vehicle – bicycles for close deliveries, regular bikes for further," Ramkumar says.
Algorithms consider real-time traffic, rider familiarity, and vehicle load to assign the most efficient routes. "It’s an exciting problem to deliver in 10 minutes while keeping riders safe," Daga says, stressing the importance of rider safety.
AI operates in parallel across the user interface, dark store, and rider allocation to ensure seamless, ultra-fast delivery. "Millions of decisions are made per second to meet the 10-minute promise," Ramkumar explains.