I will be taking on a new role as a Data Scientist at Schlumberger's Software Technology and Innovation Center starting July 2019 after receiving my Ph.D. in Computing and Mathematical Sciences from California Institute of Technology, where I was also awarded the Amori Outstanding Dissertation Prize for my Ph.D. thesis on "Online Platforms for Networked Markets".

Through my Ph.D. studies, I was supervised at Caltech by Prof. Adam Wierman and Steven Low, and funded by Singapore through the National Science Scholarship. My main research interests lie in the intersection of applied mathematics, economics and computer science.

To find out more about me, check out my work experience, my research summary and papers, or download my curriculum vitae here.


I am excited to be joining Schlumberger's Software, Technology and Innovation Center (STIC) in Menlo Park as a Data Scientist starting July 2019.

During my time at Caltech, I was fortunate to also spend two summers at STIC, where I help automate processes in Schlumberger, using a combination of machine learning, reinforcement learning, signal processing and optimization.

Prior to my Ph.D at Caltech, I worked as a research engineer at the Institute of High Performance Computing, one of Singapore's national research institutes. During the nine-month stint, I was involved with two projects, which led to multiple publications.


I have published work in various topices, ranging from networked economics and platform design to online optimization. My published papers can be found on my google scholar page here, and a short abstract is included for each project below.


Platforms in Networked Markets

[PFLW, Infocom'17]: Showed that open access platform designs under a networked Cournot competition model preserve large proportions of optimal social welfare in the worst case, in stark contrast to the unbounded worst case loss in designs that control over allocations.

[LPBW, CDC'17]: Further refinements in bounds over the worst case efficiency loss in access control designs.

[PLFBW, ArXiV'19]: Journal version combining the two previous papers, with extensions of results for controlled allocation platforms to a family of objective functions, and the introduction and analysis of a search costs model.

Load-side Frequency Control

Online Constrained Optimization

Battery Swapping Algorithms


I am available by email at (firstname)pzf@gmail.com. If you would like to know more about me, you can also check out my social media accounts in the links below:

Retaliating against colluding drivers in Uber

When pushed against the wall, drivers on ridesharing applications like Uber can work together and collude, allowing them sufficient power and information to take advantage of ridesharing companies and consumers. New research at Caltech seeks to find ways of combatting this problem.

Many startups today, including ridesharing companies like Uber and Lyft, revolve around a central planner or intermediary, known also as an online platform. These platforms, do not usually themselves produce nor consume, and only serve to bring multiple parties together, such as drivers and riders on Uber. There are many ways to design these platforms, varying over a range of control and transparency.

Figure 1: Examples of platforms and the participants they bring together.

These participants do not always react to the designs the way the central planner hopes, and they may at times collude and resist the control of the platform. For example, Uber and Lyft drivers that were unhappy with the amount of pay they received recently logged off from their applications in a coordinated fashion to create an artificial demand-supply shortfall, manipulating surge pricing. The same idea of artificial shortfall of supply by Enron also caused the California Energy Crisis in 2000, costing $40B in losses. This means that platform designs need to ensure participants’ selfish actions do not harm the overall system drastically, and is an active topic of research in Adam Wierman's lab in the Computing and Mathematical Sciences Department at Caltech.

Different platform designs lead to different outcomes and can be vulnerable to varying levels of manipulation, depending on factors including the number of participants on each side of the market and non-monetary costs such as consumer search costs.

The main challenge in understanding platform designs is to incorporate potentially individualistic and selfish actions of participants, and the possibility of collusion amongst participants. Platforms such as Uber may assume that due to their scale, participants have limited information and power for manipulating prices. A disregard of the potential for collusion allowed Uber drivers to coordinate, especially near hotspots such as airports. The coordination allows them to artificially create supply-demand shortfalls and force the platform to trigger surge pricing.

Graduate student John Pang, together with collaborators at Cornell University, and his advisor Adam Wierman, study platform design under a networked economic model. In a recent publication, they showed that platforms that control allocations, when misaligned with producer incentives, can instead disincentivize production (or hours spent driving for Uber drivers). Such designs shift competition between producers to a setting where producers are incentivized to cooperate and together compete against the platform. In the same work, Pang and his co-authors showed that designs allowing access for every firm to every consumer increase competition between them and result in an “almost efficient” market, i.e., one that preserves at least a good proportion of the social welfare under a benevolent dictatorship.

“The network competition model we study had surprising results to what you would see in these platforms, such as the collusion and cooperating between firms,” says Pang. “The question now, is, can we find a solution using these models to reduce the impact and incentives of collaboration and collusion in these platforms?”

“We are now working on generalizing the work to also look at electricity markets, where such collusions manifest with much more devastating effects, such as in the California Energy Crisis in 2000,” Pang explains while introducing a new collaborator in this extension in Subhonmesh Bose, a previous student from the lab who is now an assistant professor at University of Illinois. They found that consumers can help prevent manipulation in these platforms by being flexible in their demand.

Figure 2: Drivers may collude to obtain more "market power" allowing them to make better joint decisions that hurt the system, but rewarding riders to "be flexible" with their demand reduces the impact and incentives of these collusions, as they shift "market power" back to the platform.

The first paper titled “Transparency and Control in Platforms for Networked Markets” is under submission to Operations Research, while the second paper titled “Anticipation and Demand Management in Platforms for Networked Markets” is under submission to Management Science. Both papers are currently available on ArXiV.



This is bold and this is strong. This is italic and this is emphasized. This is superscript text and this is subscript text. This is underlined and this is code: for (;;) { ... }. Finally, this is a link.

Heading Level 2

Heading Level 3

Heading Level 4

Heading Level 5
Heading Level 6


Fringilla nisl. Donec accumsan interdum nisi, quis tincidunt felis sagittis eget tempus euismod. Vestibulum ante ipsum primis in faucibus vestibulum. Blandit adipiscing eu felis iaculis volutpat ac adipiscing accumsan faucibus. Vestibulum ante ipsum primis in faucibus lorem ipsum dolor sit amet nullam adipiscing eu felis.


i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;

print 'It took ' + i + ' iterations to sort the deck.';



  • Dolor pulvinar etiam.
  • Sagittis adipiscing.
  • Felis enim feugiat.


  • Dolor pulvinar etiam.
  • Sagittis adipiscing.
  • Felis enim feugiat.


  1. Dolor pulvinar etiam.
  2. Etiam vel felis viverra.
  3. Felis enim feugiat.
  4. Dolor pulvinar etiam.
  5. Etiam vel felis lorem.
  6. Felis enim et feugiat.





Name Description Price
Item One Ante turpis integer aliquet porttitor. 29.99
Item Two Vis ac commodo adipiscing arcu aliquet. 19.99
Item Three Morbi faucibus arcu accumsan lorem. 29.99
Item Four Vitae integer tempus condimentum. 19.99
Item Five Ante turpis integer aliquet porttitor. 29.99


Name Description Price
Item One Ante turpis integer aliquet porttitor. 29.99
Item Two Vis ac commodo adipiscing arcu aliquet. 19.99
Item Three Morbi faucibus arcu accumsan lorem. 29.99
Item Four Vitae integer tempus condimentum. 19.99
Item Five Ante turpis integer aliquet porttitor. 29.99


  • Disabled
  • Disabled