About
With over nine years of experience in software engineering, I have focused on distributed…
Activity
3K followers
-
Tushar Goyal liked thisTushar Goyal liked this🎉 Today marks my 497th day at Shopify. Could I have waited three more days for a nice, round number? Yes. But that feels far too reasonable. My second Shopify Summit began at the airport at 11 a.m. I arrived at my hotel at 11 p.m. It was a one-hour flight. Thankfully, several other Shopifolk were stuck there too, so at some point we stopped being delayed travelers and became an airport support group. ✈️ Once we finally made it, Summit was amazing. I got to build, connect with incredible talent, and meet engineers I usually interact with by repeatedly appearing in their Slack notifications asking about interviews. It was nice to prove I’m a real person and not just a recruiting bot with unusually specific follow-up questions. I also met some of our Shopify Rebellion players, which the gamer in me handled with a completely normal and professional amount of excitement. 🎮 Oh, and Diplo performed. Just a very normal thing that happened at work. I finally got to meet more of my Engineering TA team in person and spend time with other TA folk from across Shopify. Turns out we’re just as capable of talking about recruiting outside of Slack. We did discuss other things too. I think. One of the best parts? Meeting so many people I’ve helped bring into Shopify. I’m sharing photos with only a small group of them, but I got to spend time with intern engineers, Apprentice Product Managers, and full-time engineers across almost every scope. Everything they’ve accomplished since joining is theirs. They found their footing, built relationships, did great work, and made their own place at Shopify. But knowing I played a part in them being here? Pretty damn cool. I left Summit with stronger relationships, new friends, a full camera roll, and even more appreciation for the ridiculously talented humans who make Shopify what it is. Day 497. What a ride 🚀 P.S. Diplo. Shopify Rebellion. In case I somehow failed to make those parts clear.
-
Tushar Goyal liked thisTushar Goyal liked thisGSL 💙 Iceberg Summit 2026 Research, engineering, open source, and community. These all go hand in hand at Gray Systems Lab. Beyond academic research, our team actively contributes to open-source projects, builds production-grade systems, and shares experiences with the broader data community. Today we are happy to share a couple of presentations we gave on Iceberg Summit 2026. We hope you find them useful and inspiring! - Column Families in Iceberg: Enabling Efficient Feature Updates for Wide Tables https://lnkd.in/gznBDMNn Gabor Kaszab, Péter Váry - From Drift to Drive: Slicing Through Data with Iceberg and Vortex https://lnkd.in/gqHsRey9 Ashvin A., Will Manning - Advancing Query Performance with Iceberg Secondary Indexes https://lnkd.in/gnT6Y5rJ Huaxin Gao, Péter Váry - Unlocking Fast, Flexible, and Future-Ready Vector Search in Iceberg with ANN Indexing https://lnkd.in/grJFkm_B Ashvin A., Suhas Jayaram Subramanya - Freshness-Aware Table Loading on the Client-Side of the REST Catalog https://lnkd.in/gEQj8ezx Gabor KaszabFreshness-Aware Table Loading on the Client-Side of the REST CatalogFreshness-Aware Table Loading on the Client-Side of the REST Catalog
-
Tushar Goyal liked thisTushar Goyal liked thisAfter 25 wonderful years at Microsoft, Diego has left the building.
-
Tushar Goyal liked thisProud to have contributed to this multi-team effort at Microsoft to bring GPT-5.6 to Microsoft 365 Copilot. Exciting to see it ship and reach customers at scale!Tushar Goyal liked thisSuper to see GPT-5.6 with Work IQ come to Copilot Chat, Cowork, M365 apps, GitHub, and Foundry today. From multi-step agentic work to analysis and content creation, it brings stronger reasoning and higher-quality outputs without sacrificing efficiency. Try it out: https://lnkd.in/giKyzdFt
-
Tushar Goyal liked thisTushar Goyal liked thisNew in-depth blog post time: "Inside TPU and GPU Clusters: The Anatomy of Collective Communication". If you want to deeply understand the core primitives behind scaling the training / inference for MoEs and dense transformers, going a level below FSDP, expert parallelism, data parallelism, model/tensor parallelism this might be a fun read. Blog: https://lnkd.in/gU3XBKvm I cover: * TPU cluster topology: (super)pods, slices, DCN, PCIe, ICI * All-Gather: 1D/2D rings, and path algo (lots of visuals so should be crystal clear how these work even if you're not a perf engineer) * Reduce-Scatter (which is the dual of AG) and All-Reduce * All-to-All (used to dispatch tokens to target experts in MoEs) * NVIDIA GPU cluster topology (reference DGX architecture): nodes, scalable units, fat tree * GPU collectives within the node: rings, trees (log2 steps), and SHARP (in network compute unit) * GPU collectives across nodes, hierarchical algorithms over InfiniBand etc. I was heavily inspired to do this deep dive after reading the excellent Scaling book by an excellent group of people Jacob Austin, Sholto Douglas, Reiner Pope, and others. What originally started as "let me maybe just make four figures covering All-Gather, Reduce-Scatter, All-Reduce, and All-to-All so I can understand them better, it shouldn't take more than a day, right, right?" somehow turned into this 40 figures later. Along the way, I realized that the collective algorithms only really make sense once you understand the underlying hardware topology. TPUs were a bit easier to reason about, but I couldn't skip GPUs, I love them too much. Rings are cool, but I also wanted to understand tree algorithms. But also SHARP, and fat trees, and hierarchical collectives. :') So the scope slowly expanded, and little by little, this blog post came to fruition. Just a side-quest. Hope you like it! :) --- Also a big thank you to my friends for reviewing the blog and providing feedback: * Aroun Demeure (prev GPU/AI stuff at Magic, GPU architect at Apple and Imagine, my llm.c buddy!) * Axel Feldman (making GPUs go brrr at Jane Street, we met for the first time at Mark Saroufim's excellent GPU mode event) * Pranjal Shankhdhar (ex xAI GPU wizard, one of two people who inspired my original matmul blog!)
-
Tushar Goyal liked thisI am absolutely thrilled to share that I’ve been selected as a finalist for the Technologist Visionary Award in the 2026 #empowHERaccess Global Prestige Awards. Building and scaling highly available data stores at Netflix is deeply rewarding work, and it's a privilege to be recognized for solving complex challenges alongside so many phenomenal leaders. Thank you, Women in Cloud, for championing women in tech, and congratulations to all the visionary finalists on this list! 🚀☁️ #WomenInTech #DataEngineering #WomenInCloud #EngineeringLeadership.Tushar Goyal liked this🎉 The finalists are here. 🎉 Congratulations to the extraordinary founders, executives, technologists, mentors, allies, and ecosystem builders selected as finalists for the #empowHERaccess Global Prestige Awards 2026. Your leadership is expanding economic access, accelerating innovation, and shaping the future of the AI economy. Join us as we celebrate this year's finalists at the Global Prestige Awards Ceremony & Dinner on July 17, then continue the momentum at the ICONIC Leadership Workshop on July 18. ✨ Meet the finalists and reserve your seat today. ✨ #empowHERaccess2026 #WomenInCloud #AILeadership #WomenInTech #Innovation #IconicLeadership#empowHERaccess Global Prestige Awards 2026: Announcing This Year's Finalists#empowHERaccess Global Prestige Awards 2026: Announcing This Year's Finalists
-
Tushar Goyal liked thisTushar Goyal liked thisTime to call it a career. I'm accepting the voluntary early retirement from Microsoft. Obligatory badge photo, along with my crystals, ship-its, and a few perks+ purchases in the background.
-
Tushar Goyal liked thisTushar Goyal liked thisAlso: Modular is still open sourcing Mojo this year, right on track. We can’t wait to share the details at ModCon, August 18th in San Francisco. https://lnkd.in/guHpEw5K
-
Tushar Goyal liked thisTushar Goyal liked thisHey, just a reminder to all you folks that got laid off recently (my ServiceNow and LinkedIn folks especially): The tech industry will grind you down and spit you out if you let it. And it’s not always a meritocracy despite everyone’s insistence- more decisions that you realize are made by higher ups and middle managers trying to cater to them. Getting laid off, reduction in forced, restructured, fired or whatever the word is does not always mean there’s something wrong with you. It might not have been a good fit with your higher ups. You might have gotten caught up by a McKinsey associate’s spreadsheet algorithm. Maybe a boss somewhere just wanted to “try something new and fresh.” Try not to take it personally- like one of comments say, it’s hard to do. But try anyways. Some of my best work and the best leaders I met (shout out to Heath Ewer) is after getting laid off. If your financial situation allows, maybe take some time off and spend it with your family and friends. You’ll never get a better time.
Experience & Education
-
Microsoft
View Tushar’s full experience
See their title, tenure and more.
or
Publications
-
CANvas: Fast and Inexpensive Automotive Network Mapping
Usenix Security 2019
-
Preventing Phishing Attacks: A Novel Approach
International Journal of Computer Applications
Patents
Courses
-
Applied Information Assurance
14-761
-
Computer Architecture
15-740
-
Computer Networks
14-740
-
Disributed Systems
15-640
-
Embedded Systems
14-642
-
Introduction to Computer Systems
15-513
-
Introduction to Information Security
18-630
-
Network Security
18-731
-
Parallel Computer Architecture and Programming
15-618
-
Search Engines
11-642
-
Storage Systems
15-746
-
Wireless Sensor Networks
18-748
Recommendations received
1 person has recommended Tushar
Join now to viewOther similar profiles
Explore more posts
-
Robert Hundt
We are reimagining PyTorch on TPU (https://lnkd.in/gCrM-TZD). This approach still calls XLA but in a profoundly different way, enabling a solid eager mode. In the limit, this new approach will enable a true JIT approach with recompilation and re-optimization, online feedback-directed optimizations, avoidance of OOMs, support for HW fungibility, hyper-optimizations of kernels, and more - all with close to zero code changes!
2 Comments -
MobiusEngine
Hardware engineers right now: business as usual. Software engineers right now: refreshing LinkedIn every 30 minutes. AI can generate code. AI can't manufacture chips. The irony: the people building AI infrastructure have job security. The people using AI tools to be more productive are getting cut. Physical constraints protect jobs. Digital abundance doesn't. The market is telling you where to invest your skills. Visibility matters. Relationships matter. Having options before you need them matters most. Source: https://lnkd.in/gahzATXv
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content