Thrilled to share that my paper: "Sprinter: Speeding Up High-Fidelity Crawling of the Modern Web" got accepted at NSDI'24. Sprinter improves crawling throughput of distributed web crawlers by 5x while ensuring near-perfect fidelity of resource fetches.

I'll be at NSDI'25 next week, chairing the Infra for ML session. Also, my team at Meta is hiring (. Please reach out if you think you'd be a good fit. Looking forward to connecting with everyone at the conference. #NSDI25 #MetaCareers #HiringOpportunity.

RT @punwitter: @unixterminal NDSL @HPE_Labs researchers @lianjiecao and @goelayu_sh will be attending #MLSys as well. BTW, drop by labs if….

I will be at @usenix NSDI this week presenting our work Sprinter (. Happy to chat with graduating students to share my experience navigating the job market as a systems PhD. #nsdi24.

RT @what_007: We are working towards using this system to fix tons of broken citations on Wikipedia. Excited about translating our research….

Also I am on the job market for industry research and academic postdoc positions. If you are hiring this cycle, pl reach out to goelayu@umich.edu.

This paper was the last chapter of my thesis and couldn't have expected it to be accepted at a better venue. The preprint will be out soon.

RT @ymanerkar: I am recruiting PhD students! If you're interested in the intersection of formal methods and computer architecture/systems r….

I'll be at UT Austin this Monday presenting my work. How plausible does the networking community think reinventing the web is, to combat such problems? Note that founders of the web proposed bi-directional links as a potential solution, but it has its own shortcomings!.

Link to the paper:

Links on the web suffer from poor resilience, ever since the advent of WWW. Efforts that involving rearchitecting the web, for e.g., content based networking (IPFS) have shown poor adoption. Our HotNets'22 paper addresses these concerns and paves the path to a less ephemeral web

I will be presenting my work Jawa: Web Archival in the Era of JavaScript at OSDI'22 on Wednesday, 13th July, 4:20pm PDT 📢📢📢. Please drop by if you are curious to learn about the various insights we drew from program analysis of JavaScript in an archival context. #osdi

Are there any well-known DL models/classifiers that can be used for computing semantic similarity b/w two pieces of text? In other words, ability to determine if a particular update to a text alters its semantic meaning.

Registered for OSDI! If anyone is still looking for roommates, kindly let me know.

RT @NetPreserve: ⚡️TALKS: .🟠Sanaz Baghestani| Encyclopaedia Islamica Foundation @SSharifabadi | @AlzahraUniverse .Fariborz Khosravi| @nlai_….

Systems research can taken forever to yield results but I hope this motivates others to keep persevering, and to believe in the process :).

However as of today, we have leveraged (and tweaked) that same tool to help us build three systems and publish them at OSDI'21 (hocrux), OSDI'22 (Jawa) and HotMobile'21.

Went through a rollercoaster of emotions, losing hope given the extent of JS features that need to be handled, and the notorious dynamic nature of JS resulting in new bugs every day. Eventually we decided to pivot to other things given the lack of promising results.

I spent the first two years of my PhD doing absolutely nothing but building a JS analysis framework for a web performance optimization we were trying to implement.

while ensuring high page fidelity. I would like to use this opportunity to share a small story about "hope" from the last few years of my PhD that hopefully resonates and/or inspires other students. 2/7.