ORIE Ph.D. interns help eBay deliver “Now”

While Amazon CEO’s Jeff Bezos got extensive publicity by announcing a future drone delivery service on 60 Minutes, Amazon, eBay and others already offer more down-to-earth rapid delivery services using a mix of bicycle couriers, vans, motorbikes and small  parcel cars, depending on the city. 

ORIE Ph.D. students Patrick Steele and Chaoxu Tong have spent the past semester at eBay developing  an array of operations research tools to reduce the cost of rapid delivery by eBay’s recently acquired company, Shutl.  They are among several ORIE Ph.D. students involved in internships this year.

“I think this experience  will broaden my scope on research problems I plan to tackle later and makes me think more about how we could apply theoretical techniques to real problems more efficiently,” said Tong.  “The most striking thing I learned is that a significant amount of effort may be needed to work with real world data, which is both rich and messy.  Another thing is the delicate balance between a tractable model and reasonable assumptions.” 

Shutl provides fast (typically one-hour) delivery of products from local stores, and is now offering those services for products bought on eBay in some select cities.  These courier deliveries can also be scheduled in any future one-hour delivery.   EBay also offers a companion service, eBay Now, in which a ‘valet’ is dispatched to go into stores and purchase products from the shelf for a customer.  While Steele and Tong are concentrating on Shutl, they anticipate that their work will apply to eBay Now as well. 

Reducing operational costs

Steele and Tong are working on reducing operational costs for Shutl via more efficient allocation of pickup and delivery jobs to couriers and scheduling of those jobs for each courier.  A single courier needs to complete a number of jobs, each with specified pickup and delivery locations and time windows.  “This leads to an interesting scheduling problem,” said Steele.  “The problem grows far more complex when you consider partitioning jobs among a collection of couriers, and even more so when you need to anticipate jobs arriving beyond the current time horizon,” he noted. 

The work of the Ph.D. interns intersects ORIE’s curriculum in several ways: 

• They use statistical techniques to analyze demand patterns and courier behavior, creating models for predicting delivery success and failure.  
• They employ various algorithmic techniques in combinatorial optimization to create models and heuristics to solve the hard scheduling problems, including approximation algorithms and related theory. (“It would be nice if we can construct lower bounds on solution quality, which would enable us to know where the sky is while we try jumping higher,” said Tong).  
• And the pair uses computer simulation to see how their models and heuristics scale with demand levels for which they do not yet have data, such as when starting operations in a new city. 

A nice contrast

Steele spent 2 ½ years in Ithaca and Tong spent 3 ½ years there before taking up the internship.  “The work provides a nice contrast to the theory work I have done at Cornell,”  says Steele, “and lines up nicely with problems I have studied there.  The most striking aspect of the work is the complexity of reality: most models assume highly idealized scenarios that don’t apply to the exigent situations which arise in practice, such as inclement weather, mechanical failures, and concern for customer satisfaction.  I think my experience here has offered a useful perspective on how theory works in practice that will help me in the future in both industry and academics,” he said.

As part of their research at eBay,  Steele and Tong, who are working in eBay offices in Bellevue, WA, “did a ride-along in San Jose CA with an eBay Now valet,” had conversations with several courier firms in Manhattan, and are working with two eBay employees from Shutl who are located in London.  In addition, “we have a chance to work with Kamal Jain, an excellent researcher on approximation algorithms,” said Tong.  Jain, a successful Microsoft researcher who has published dozens of papers, holds dozens of patents, and is now at eBay, “enjoys mentoring [and] has mentored dozens of PhD students, most of whom are now at top universities and companies,” states eBay’s web site.

Both Steele and Tong are Ph.D. students of ORIE Professor and Director David Shmoys, who says that internships provide a natural first point of contact with a company – but that the hope is that further projects, including joint faculty-eBay research projects, will evolve from this first step.  

Other interns

While Steele and Tong have been at EBay since the beginning of the year, other ORIE Ph.D. students are are embarking  on summer internships. 

• Anton Braverman is a trading intern at Susquehanna International Group in Philadelphia.
• James Davis and Jacob Feldman are at Amazon in Seattle, WA.   
• Daniel Fleischman is at the Silicon Valley Center of Microsoft Research, where he is working on large-scale network optimization.
• Jiayang (Angela) Gao is interning at SAS in Cary, NC, where she is modelling how best to display products in a store or web site and at what price.
• Jiayi Guo is an intern at Argonne National Laboratory.
• Nanjing Jian is a supply chain analyst for Novartis-Sandoz in Princeton, NJ.  She is working on the allocation of orders in their supply distribution centers. 
• Eric Ni is working as an intern in Goldman Sachs’ equity strategy group in London. 
• Alice Paul is an intern in the complex systems engineering group at the GE global research center in Niskayuna, NY.