$49.5M
Cost Savings
$12M
Revenue Generated
9.28%
Late Delivery Reduction
6 years
PM + Engineering
Featured Work
Highlights from building ML-powered products at scale
ETA-Dispatch Coordination
Grubhub
Coordinated ML-generated ETA predictions with dispatch optimization to improve bundled delivery economics
Priority Delivery
Grubhub
Premium delivery tier balancing dispatch constraints with user experience — 550K+ orders, 46% adoption
ETA Display Experimentation
Grubhub
Designed for ML uncertainty — wider ETA ranges improved outcomes without hurting conversion
Improving Delivery Time Predictions
Grubhub
Diagnosed ground truth data quality issues in food prep predictions and redesigned the modeling approach
Building
From transformer architectures to AI communities
Efficient Attention in Vision Transformers
Benchmarked efficient attention (Linformer, Performer, Nyströmformer) for ViTs in PyTorch. Proposed hybrid with improved accuracy-compute tradeoffs.
No-Code Works
Founded Cornell Tech's AI builder community. Teaching 90+ students to ship products using Claude Code, Cursor, and more.
Village
Cross-platform childcare app built in 4 weeks with AI-native tools. Full product lifecycle from research to production.
About
Pranav Dhingra is a product manager and engineer who builds ML-powered products at scale.
At Grubhub, he led timing and estimation systems serving millions of deliveries, coordinating across ML, dispatch algorithms, and experimentation platforms. His work on ETA optimization generated $49.5M in annual savings.
He's currently pursuing an M.Eng. in Computer Science at Cornell Tech, focused on LLMs, deep learning, and reinforcement learning. He founded No-Code Works, teaching 90+ students to build products with AI tools like Claude Code and Cursor.
Previously, he was a software engineer at Grubhub, building the switchback experimentation platform and maintaining auto-dispatch systems for 500K+ daily deliveries.