Milind Chabbi '15 continues compiler, HPC research in industry

Rice computer science alumnus communicates research opportunities and results at Uber.

Headshot of Milind Chabbi

Although his Rice University Ph.D. advisor, John Mellor-Crummey, is focused on performance tuning for a national exascale research computer, computer science alumnus Milind Chabbi (’15) prefers solving high-performance computing (HPC) challenges in the technology industry. For the last five years, his research has supported Uber’s mobile and microservice-based programming environments. He has applied his learnings in performance and correctness tools to solve large-scale efficiency and reliability problems at Uber.

“When I came to Rice, I already had exposure to industry –how it works, how research is applied, and the kinds of limitations research often hits in industrial setting,” said Chabbi. “This helped me identify and focus on problems that are intellectual in nature while also having a mindset that enabled me to envision how the results of that research could be applied to real-world problems.

Learn to communicate about interdisciplinary research 

“After my Ph.D., I really learned to appreciate the interdisciplinary nature of research —where we identify initial solutions for a complex problem, then discover the problem is presenting itself in slightly different forms across different domains. Can we apply our original solution directly in those domains or do we need to modify the solution to better fit the newer and changed environments?”

His interest in bridging the gap between software applications and computer architecture requires Chabbi to communicate with a wide variety of engineers and stakeholders. In addition to identifying and resolving HPC, compiler, and other performance and deep learning issues, he also promotes research as an important part of team culture.

Demonstrate the value of research to leadership

Chabbi said, “Each of my industry roles has been research focused, and I try to set up or expand the research mentality among the leaders with whom I interact. In large organizations, there are different teams and projects; not everything is research focused. I love to demonstrate how research can overlap and benefit areas that haven’t previously considered that aspect of problem solving. If I join an organization that may not have had a research focus when it was formed, then I try to bring it with me.”

To raise awareness about the benefits research can bring to a commercial tech company, Chabbi organizes reading groups in which his colleagues can discuss relevant papers. He champions the hiring of Ph.D. interns and works with them to publish papers about new findings that benefit both the organization and the scientific community. He also encourages his peers and interns to join him in attending conferences where new research is being presented.

“One tactic I use for introducing research into our process is by looking for problems that allow me to take on longer-term projects, particularly ones that have a sense of uncertainty or lack a clear-cut solution but may solve more complex problems in the long run,” said Chabbi.

Include the ‘why,’ especially when pitching a project to management

Gathering data and demonstrating the recurrence of the problem is often his first step in pitching a new proposal to management. Making a clear statement about an issue should always include a ‘why’ for management to grasp. For example, Chabbi might point out that engineering resources currently spent addressing a recurring problem could be redirected if the underlying issue were resolved.

He said, “Another way to attract interest in research proposals is to start a project through interns or university collaborations. Showing the value of the work while keeping a low initial investment is how one of our recent solutions began. Working with Georgian-Vlad Saioc, my 2022 Ph.D. intern from Aarhus University, we devised a tool for detecting goroutine leaks. The moment we started working on it, we realized the value it could bring and were able to build a case for investing additional resources and involving more and more teams. We delivered a robust solution – LeakProf – that is now a pillar of the work we do in our team.”

LeakProf, like other industry solutions Chabbi has helped create, is an example of long-term projects that revolve through a cycle of problem identification, clarification, research, and application, then repeat the process. He believes research flows beyond traditional software and hardware boundaries and said a typical research-based problem can’t be compartmentalized as ‘my work’ versus ‘your work.’

“The appeal of that interdisciplinary aspect of research in industry is offset by uncertainty. Researchers have to be comfortable living with many unknowns. And that is where communication becomes critical. Despite the unknowns, you’ll have to talk about what you are working on and what you need –particularly from managers and leadership. And you must communicate the needs in a way that keeps everyone excited about the research and its possible application or outcome.

Tell your story to different audiences

“Learning to tell stories to different audiences is one of the skills we learned and practiced at Rice, thanks to Tracy Volz and the ACTIVATE Engineering Communication Program,” said Chabbi.

Chabbi was an experienced Ph.D. student when he met Volz during a welcome dinner for incoming graduate students. After the meeting, he contacted her about practicing his presentation for an upcoming talk.

“I knew from that rehearsal that Milind would make a good coach,” said Volz, who invited him to work with other graduate students on their presentations.

Chabbi was flattered and accepted the invitation. He began coaching his peers and continued seeking assistance from Volz and her team to improve his own technical communication skills.

Coach and create more research storytellers 

“Using their input to refine my slides and pace my talking speed was very helpful,” he said. “Soon I was doing a decent job in technical presentations and at one conference talk, I won a presentation award.

“It was exciting to work with other Ph.D.’s and see their conference talks improve; I wanted every Rice student to give an excellent talk. I really enjoyed coaching and continue to coach my interns in their talks and presentations today.”

In addition to coaching his interns, Chabbi leverages his communication skill to tell the story of his research team’s opportunities both up and down the chain of command.  He said communicating upwards may prompt him to simplify certain aspects so that leaders can quickly grasp the information they require to make a decision.

Find common ground with other engineers

“Talking with engineers who have expertise in different technical areas takes a slightly different approach,” Chabbi said. “They may not understand exactly what you are doing, but you need their help to solve part of the problem. Talk in their lingo! Communicating with other engineers starts with understanding their level of knowledge. You may end up teaching each other about how something works or isn’t working. Be patient in sharing your expertise; the more each party understands, the more easily you can work together to resolve the issue.

“Being able to succinctly communicate our ideas to leaders and peers is exciting, and I am fired up about publishing our research across the company. I feel just as excited about the last paper I published with a Ph.D. intern as the first Ph.D. paper I published at Rice.”

Chabbi is continually intrigued by how research happens every day and how it works to solve the problem at hand. He finds it is easy to lose all sense of time as he works and said his constant learning and applying what he learns doesn’t feel like a job. For Ph.D. students who are similarly fascinated by research and want to continue their work in industry, Chabbi recommends taking advantage of the university’s communication resources.

“Yes, your research is important,” he acknowledged, “but you must be able to tell your story and show why the work is important. Take time to practice, both before and after you receive feedback from your coaches. Watch a video of yourself giving a talk. That will really help you improve.”