Xander Chin (@XanderChin) header
Xander Chin (@XanderChin) profile picture

inference @nvidia | eng @westernu @schulichleaders | building and learning for fun

sf/to
Joined February 2015
Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)Photo by Xander Chin (@XanderChin)
hand-controlled boids
310
1,455
130
20,593
742,469
Xander Chin retweeted
looking for a winter 2027 internship! i have experience working with compilers and RTL, and more recently i have been exploring LLM inference serving and optimization. would love to join a team working on similar problems this winter! Canadian citizen, open to relocating. DMs open :)
17
11
3
53
3,828
Xander Chin retweeted
started messing around with an so1o1 arm and it kinda made me feel like a kid again :3
19
9
121
4,557
the only moat left is the love of the game
75
1,546
129
9,357
269,482
a few thoughts on this: - it is now easier to create more than to create less - create less, but with great intention/meaning - think about what you are putting out there and why
i think we've glorified productivity so much to where we've lost the joy of creating. i miss the old state of twitter (x) because it felt raw, and nobody really gave a second thought to how well their thoughts would perform online, (speaking purely in terms of engagement). though hacker news and hackaday etc... still preserve remnants of this vanishing culture
2
18
784
Xander Chin retweeted
A year ago, I entered the world of robotics through a hackathon that me and @librendaa072 did. despite knowing nothing, i enjoyed the experience of building a hardware project immensely more than any of my software ones (immediately regretted picking CS instead of Tron) after one hardware hack came another, and then another. All while me and Brenda hosted events across Canada and the US. our revelation? we can’t wait for humanoids, but we also couldn’t imagine them—in their current state, smoothly integrating into society. perhaps this can be solved in the way we know best: events... but this time, to *also* capture social data. want a spot at our next event? leave a comment or DM us :)
What used to be sci-fi a year ago is swiftly becoming a reality. Robotics companies around us are racing to create the first, deployable robot into human society. Training is continuously evolving in the age of physical AI, and it won’t be long before clanker friends join us in our homes, in our hospitals, and everyday life. In 2035, 13M of walking robots in society are expected to be humanoids. It’s time to teach these robots how to play, and interact in our society. Introducing: Interplay For years we’ve hosted events for the people. Somewhere along the way, startups and companies picked up on going IRL as the new method for building trust, creating community, recruiting, and advertising products. Interplay is creating an entirely new category of events: those curated to bring out the best in humanity… carefully designed to train the best social robots. Want to teach them? Leave a comment or DM us to grab a spot at our next event :)
11
4
1
30
10,440
Xander Chin retweeted
we made a @bracketbot play cup pong! (or at least, we tried) first time pulling an all-nighter and learning about how these cuties worked!
13
3
1
97
5,987
Xander Chin retweeted
you seek more intelligence, Arjuna, when what you lack is clarity
31
475
28
5,347
77,207
check it out
we didn't know how to design a chip a couple months ago so we reinvented Groq's LPU to run inference on @karpathy 's MicroGPT! here's a breakdown of what we learned:
1
17
1,824
if you are ever looking for some guidance hit up @aatikchopra. he’s one of the best coaches out there. Aatik has helped me understand myself and guided me to solve my problems rather than just giving me the answers. you won't meet another coach who has the same level of curiosity, compassion and investment that Aatik has for you
3
1
14
1,838
Xander Chin retweeted
hey @FIFAWorldCup is this valid? we just ran inference on our reinvented Groq LPU using our custom 2.5k param model!!
7
5
1
23
4,484
Xander Chin retweeted
Reinventing Groq's LPU with @sakshambatraa and @michael_trbo . Recently we transitioned from INT precision to floating points. Here's a breakdown of how we did it:
7
6
2
22
4,182
Xander Chin retweeted
The LPU team at @nvidia is building the future of low latency inference. We're going all in on Rust. If you're a cracked Rust engineer and love making hardware do things, we want you. nvidia.wd5.myworkdayjobs.com…
14
43
6
393
49,967
I dropped out of high school at 17 to build my first startup. Now, at 20, I'm thrilled to share that @TryOmenAI has raised $41.5M in total funding to build intelligence protecting the world's most critical AI compute infrastructure.
44
25
25
249
443,963
This sentence from Carl Jung hits hard. “No matter how isolated you are and how lonely you feel, if you do your work truly and conscientiously, unknown friends will come and seek you.”
51
827
16
6,028
106,492
Xander Chin retweeted
Want to see distributed computing explained via Pong? Inspired by TinyTPU and TinyTapeout workshop at FOSSi, I wrote a paper under a week pairs this demo with a proposed next-gen optical I/O chip architecture & a roadmap to prototype it. Read it on GitHub: github.com/llhtimlam/tt_um_l…
7
10
1
106
22,471
Xander Chin retweeted
A few months ago, I saw Karpathy build NanoChat in PyTorch, and it made me want to understand how these models work underneath the abstractions. So I decided to try building one myself, but in a different framework: JAX. Here’s how I did it: 🧵
5
7
25
1,901
Xander Chin retweeted
reinventing Groq's LPU with @michael_trbo we got instruction driven data movement working between SRAM memory blocks and MXM compute!!
11
18
1
76
7,028
Xander Chin retweeted
We implemented @karpathy 's MicroGPT fully on FPGA fabric. No GPU. No PyTorch. No CPU inference loop. Just a transformer burned into hardware, generating 50,000+ tokens/sec. The model is small, but the idea is not: inference does not have to live only in software 👇
260
681
121
7,386
857,532