Ramya Rajasekaran
San Francisco, California, United States
680 followers
500+ connections
About
Hello! I'm Ramya, a software engineer looking for exciting challenges and interesting…
Activity
680 followers
-
Ramya Rajasekaran shared thishttps://goo.gl/Mfu2ib Earn a chance to go to GHC this year!
-
Ramya Rajasekaran reacted on thisRamya Rajasekaran reacted on thisAfter 10 years and one day, today is my last day at Salesforce. I'm grateful for the opportunity to write parts of the company’s story and for my part in the stories of the people I worked with. As for what’s next, I'm taking an extended break from full-time work, something I haven't done in twenty years. Time is a resource to spend mindfully, and I plan to double-down on the things that I enjoy — family and friends, skiing and climbing, building things, and connecting with others. I've budgeted some time to rejoin the technology and entrepreneurial community at large, so drop me a line if you'd like to tell me a story over a cup of coffee or connect me with someone. And for those of you who were hoping for some wit and humor, perhaps an extended riff on an analogy like "Unless you're the lead sled dog, the view never changes" or The Major General's Song remade along the lines of "I'm the very model of a lower upper manager", rest assured that those are still available in person.
-
Ramya Rajasekaran reacted on thisRamya Rajasekaran reacted on this💸 Keep a closer eye on your money with Digital Wallet for Data Cloud. Make better budgeting and buying decisions by tracking and identifying trends in your product consumption in near real-time. sforce.co/4694eMt
-
Ramya Rajasekaran reacted on thisIt’s been such a fun ride working on this product with an incredible group of PMs, UX and Engineers to bring this from concept to reality! I’m so excited to get Digital Wallet into the hands of our customers and continue to expand it to more products 🤩Ramya Rajasekaran reacted on this📣 Digital Wallet for Data Cloud is officially here. Make better budgeting and buying decisions by tracking and identifying trends in your product consumption in near real-time. sforce.co/4694eMt
-
Ramya Rajasekaran reacted on thisChandra Family Department of Electrical and Computer Engineering (Texas ECE - UT Austin ECE)
Chandra Family Department of Electrical and Computer Engineering (Texas ECE - UT Austin ECE)
Ramya Rajasekaran reacted on thisThis fall, we will say goodbye to Prof. Christine Julien who will be leaving the Chandra Family Department of Electrical and Computer Engineering (Texas ECE - UT Austin ECE) to become chair of the Virginia Tech Department of Computer Science. For 20 years, Prof. Julien has been an indispensable part of the Texas ECE community. Her leadership, dedication, and passion for education have influenced and inspired countless students and colleagues. Thank you, Prof. Julien, for everything you have done for Texas ECE, and we wish you all the best in your new position. -
Ramya Rajasekaran liked thisRamya Rajasekaran liked thisIt's finally here, my print copy of AI at the Edge! 😁 Get yours today! https://lnkd.in/ebcV_iWa #aiattheedge #ai O'Reilly Media Tim O'Reilly Edge Impulse
-
Ramya Rajasekaran liked thisRamya Rajasekaran liked this🚨🚨 Attentive & Salesforce 🚨🚨 Super excited to share the news that the Attentive and Salesforce Commerce Cloud integration is live! 🔥🔥🔥 This launch marks a significant expansion to our partnership and compliments our previous integrations with Salesforce Marketing Cloud and Service Cloud ❤️❤️❤️ Brands can now connect their customer and product data, and utilise SMS across the customer lifecycle to: 📈 Increase personalization and list growth 🤝 Deliver exceptional customer experience 🚀 Maximize campaign performance Check it out: 👇🏾👇🏾👇🏾 #smsmarketing #conversationalcommerce #beattentive #salesforce #partners
-
Ramya Rajasekaran liked thisRamya Rajasekaran liked thisAfter nearly 4 years in Search and AI at Salesforce, working with an amazing crew of people across the Pacific Northwest, SF Bay Area, France, and Israel (among other places!) working on a wide array of pretty cutting edge ML technologies and helping various teams apply them to personalized search, chat bots, recommenders, and question answering… the time has come to turn in my #badgegunandsteed My last day was this past Friday, but my final day was actually so busy that I didn’t have time to post this then, so it’ll have to go out now! Aloha, my Saleforce friends - I’m told that word can mean many things, but for now it’ll have to mean au revoir!
-
Ramya Rajasekaran reacted on thisRamya Rajasekaran reacted on thisI had them pull me an extra shot of espresso this morning to remember Vijay Gill. Direct, complex, stimulating, strong with a hint of bitter that puts the sweeter notes in context. Thank you and Godspeed, my friend.
Experience
-
-
-
-
-
-
-
-
-
Education
-
The University of Texas at Austin
-
-
Activities and Societies: Women in Electrical and Computer Engineering, Eta Kappu Nu (HKN)
-
-
-
Volunteer Experience
-
Volunteer
WECE
- 3 years 10 months
Children
K-12 Outreach
Courses
-
Algorithms
-
-
Computer Architecture
-
-
Concurrent Programming and Distributed Systems
-
-
Data Science Laboratory
-
Honors & Awards
-
University Honors
University of Texas at Austin
University Honors
-
Valedictorian
Central Board of Secondary Education, India
Recognized for reasons of outstanding academic performance and for being among the top 0.1 percent of successful candidates of All India Senior Secondary School Certificate Examination 2016. Ranked 1 out of 380 students. Received scholarship for scoring 97.4%.
Languages
-
English
Native or bilingual proficiency
-
Tamil
Native or bilingual proficiency
Organizations
-
Eta Kappa Nu (HKN)
-
-Electrical and Computer Engineering Honor Society
-
Women in Electrical and Computer Engineering
President
-- Led monthly all-group meetings, and bi-weekly leadership meetings. - Represented WECE on various campus-wide committees to advocate for equal representation. - Established and maintained 'families' to ensure a sense of community
Other similar profiles
Explore more posts
-
Shaunak J
I just published Part 2 of my series inspired by Andrej Karpathy’s idea: 👉 “LLMs are becoming the new Operating System.” In Part 1, I shared my journey through Windows, laptops, Android tablets, and how LLMs feel like the next massive shift in how we interact with computers. Today in Part 2, I’m opening up the hood and breaking down the actual architecture of the LLM OS. Karpathy visualized it with a simple diagram… and the deeper I looked at it, the more it felt like a blueprint of a new kind of computer. In this article, I break down: ⚙️ LLM as the CPU 🧠 Context window as RAM 💾 File system + embeddings as the disk 🧮 Classical tools as system calls 🎥 Video/Audio as peripheral I/O 🌐 Browsers + other LLMs as networking Everything we know from traditional computer systems is reappearing — but rebuilt around language, reasoning, and models instead of code and circuits. If you’re curious how an LLM actually functions like a computer, this part makes the picture clear. 📌 Read Part 2 here: https://lnkd.in/gDggQc52 Part 3 drops next: The Context Window — Why It’s the RAM of the LLM OS. Would love to hear your thoughts! 🚀
2 Comments -
Harshit Rathod
Your GPU is idle 99% of the time while serving LLMs. Here's why. A single LLM request is secretly TWO completely different workloads: Prefill (processing your prompt) is compute-bound — it saturates the GPU's math units. Decode (generating tokens one at a time) is memory-bandwidth-bound — every single token requires re-reading ALL model weights from memory. An H100 needs ~295 FLOPs per byte to stay busy. Decode delivers ~1. That one fact explains almost everything about modern LLM serving infrastructure. I just published Part 1 of my 5-part series on LLM inference — a tutorial-style deep dive written for backend engineers, no ML background needed: → Why prefill and decode have opposite bottlenecks (with the roofline math) → How continuous batching (Orca) killed head-of-line blocking for a 36.9x throughput gain → How PagedAttention borrowed OS virtual memory to cut KV cache waste from 60-80% to under 4% → Why production systems now run prefill and decode on SEPARATE GPU pools (DistServe, NVIDIA Dynamo, Mooncake) If you understand load balancers, paging, and connection pools, you already have the mental models — this post just maps them to GPUs. Read Part 1 here: https://lnkd.in/dRbD6p57 #SystemDesign #LLM #DistributedSystems #AIInfrastructure #BackendEngineering
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content