Adithya Abraham Philip

About Me

I am a fifth-year CSD Ph.D. student at Carnegie Mellon University, advised by Prof. Justine Sherry. I do research on computer networks, with a focus on Congestion Control Algorithms (CCAs) and their increasingly complex interactions. My current research investigates the fairness properties of CCAs in various settings and deployments, so we can make informed decisions on what CCAs we should be using and where.

Before I delved into the mysteries of the Internet, I applied Machine Learning techniques to speed up software engineering pipelines as a Research Fellow at Microsoft Research India. Our work resulted in a number of publications at top venues and won the Jay Lepreau Best Paper Award at USENIX OSDI 2018.

In my spare time I dabble in archery and fix holes in my drywall. Completely unrelated, I promise.

Publications

Work Experience

Netflix - Research Fellow/Intern | March 2024 - Present

• Reverse engineering network conditions from network traces.

Google - PhD SWE Intern | June 2021 - August 2021

• Improved the congestion control algorithm BBRv2’s RTT fairness when using ECN, allowing data-center traffic to co-exist with Wide Area Network traffic while still keeping network delays low.

Microsoft Research India - Research Fellow | July 2017 - July 2019

• Developed techniques to address a variety of issues in software engineering, including risky-change prediction, test suite minimization, and automated bug-localization.
• Reduced time taken by internal testing infrastructure by 60% using machine-learning based techniques.
• Jay Lepreau Best Paper Award at OSDI’18 for work on bug-localization, and two more publications at ICSE’19 and PROMISE’19.

Microsoft Research India - Research Intern | August 2016 - January 2017

• Interned under Ranjita Bhagwan and Saikat Guha and built a clustering-based anomaly detection system to identify atypical behavior in network resource accesses.

Goldman Sachs - Summer Technology Intern | June 2016 - July 2016

• Reduced time spent by employees on regulatory reports by 54% by developing a statistical anomaly detection and mailing system to highlight suspicious activity.