Trinadh Venna

Trinadh Venna

FeaturedCompute

Four Sparks and a Closet

Part of my job is keeping an H200 cluster alive for research projects and workloads heavy enough to need it. Eight of those GPUs wired into one NVLink domain can hold a 405 billion parameter model in memory

Published
Reading
4 minutes
Image
Official NVIDIA image of the DGX Spark
Read post

Recent writing

Compute · 5 min

Dense, MoE, and the SSE Bug That Ate My Chat Stream

I had three vllm serve processes running on the H200 cluster, one prompt piped into all three through the same OpenAI-compatible client, and I was watching what should have been identical scaffolding produce

All writing

Dense, MoE, and the SSE Bug That Ate My Chat StreamI had three vllm serve processes running on the H200 cluster, one prompt piped into all three through the same Compute 5 min
Four Sparks and a ClosetPart of my job is keeping an H200 cluster alive for research projects and workloads heavy enough to need it. Compute 4 min
The Dual Nature of Reality: Waves, Particles, and Quantum StatesHuman intuition is strictly macroscopic. We evolved in a world governed by classical mechanics. In this world, a rock is Physics 4 min
Architecting A H200 HPC Cluster for AI ResearchAI training is massively GPU intensive. Until recently, our institution's AI lab operated on dozens of standalone desktops Compute 5 min
Attention is All You Need: The Paper That Sparked a Trillion-Dollar IndustryIn the summer of 2017, a team of researchers from Google Brain and the University of Toronto published a paper Compute 2 min
Real-time Robot Control with Leap MotionUsing Leap Motion Sensor to Control a Robotic Arm in Real-time. Compute 5 min
Quantum Computing 101: A Quick Guide for Busy MindsIn the ever-evolving landscape of technology, quantum computing stands as a revolutionary force, promising to transform the way we Physics 2 min
Understanding Prompt Engineering: The Art of Guiding AI ModelsIn the realm of artificial intelligence, prompt engineering plays a pivotal role in shaping the outcomes generated by language models. As we delve into the intricate world of AI, understanding prompt engineering becomes essential for fine-tuning models and harnessing their potential effectively. Compute 4 min
A Spark and a Virus2020 has been widely and unofficially considered as the worst year in many decades. If you are reading this, you Essays 2 min
The cosmic perspective on lifeOur relationship with the cosmos is not just spiritual. But in a way, physical. We are special not because we are different from each other, but the same. Essays 1 min
Story of a Simple Blob: A Humbling Perspective on a Human Centric WorldOnce upon a time, there was a tiny, smart, lively, and lonely blob on a rock floating in space entirely Essays 3 min
Order in the Chaos: The Butterfly EffectJust a mere few thousand years ago, we looked towards the sun to know what time of the day it Essays 4 min
A Ride In The Train That Never StopsIt just feels like yesterday playing with the toys in the home and thinking that my home is everything that Essays 7 min
A Comprehensive Guide on Building a Modern WebsiteIf you are someone who is brainstorming about building a website for your business, brand, or personal project, this article Compute 8 min

New writing, by email

No schedule and no noise. Just a note when something new is published.

Check your inbox to confirm. Thank you.