Ramith Hettiarachchi
@ramith__
Followers
1K
Following
8K
Media
379
Statuses
2K
PhD Student @CMUPittCompBio / @SCSatCMU Interested in ML for science while striving for trustworthiness
Pittsburgh, PA
Joined April 2014
We’ve been making phylogenetic trees differentiable :) Check out our work at #ICML2023 workshops - Sampling and Optimization in Discrete Space (SODS) ፨ and Differentiable Almost Everything (DiffAE) 〆 Looking forward to discuss and learn more! 🙌 ❤️ work with Avi & @sokrypton
4
59
306
Update: I was wrong, building the image is easy, but when you run it, you encounter numerous issues. Here’s a summary: https://t.co/34JtZjRHnp
0
0
0
Installing alphafold3 on the Dgx spark was pretty easy, but I forgot that JAX preallocates gpu memory* (XLA_CLIENT_MEM_FRACTION=0.95) Forgot to change those settings, and the spark froze without me being able to ssh 😆 * In the spark, it’s unified memory Gotto test again
1
0
2
@deedydas 😅 all of this is true. i owe a bunch to my mentors as well. my mentor at nyu was then phd student @psermanet, one of the kindest human beings i know Yann LeCun gave me opportunity 2x when i didn't have another option forward in AI. the person who pushed me to go to IIIT for my
41
97
2K
tried setting up network file storage (NFS) for the first time 😄 kinda cool :O its capped at 1Gbps though (~100MB/s read write speed).. starting to realize the importance of copying data to a compute node when doing an optimal training run
0
0
1
We want more AI in the world. We need to be able to use and experiment with all the kinds of compute we can. We need to be able to use the fastest hardware out there! 💨
2
15
127
Messy thinker? That's a superpower! Find panel coming soon in Muse helps you jump straight to your thought in a busy board.
0
2
10
Tatta Bio began with @Micro_Yunha + @ancornman1 building genomic language models and exploring how AI could help interpret the growing volume of biological data ( https://t.co/TZN6BPLXXt). Our first tool, Gaia, introduced the idea that language models could accelerate biological
0
11
48
Oh, come on! It’s only Wednesday, and 3 AlphaFold3-like models have already been launched in this week. OpenFold3 https://t.co/WtglkE220t Perl https://t.co/VHWpS0hoUn ODesign https://t.co/VZaMd5v13n
arxiv.org
Accurately predicting the three-dimensional structures of protein-ligand complexes remains a fundamental challenge in computational drug discovery that limits the pace and success of therapeutic...
3
22
160
I was impacted by Meta layoffs today. As a Research Scientist working on LLM posttraining (reward models, DPO/GRPO) & automated evaluation pipelines, I’ve focused on understanding why/wehere models fail & how to make them better. I’m looking for opportunities; please reach out!
122
235
3K
got dgl (2.4.0) to compile on blackwell architecture.. the env: torch 2.9 + cuda 13 + dgl 2.4.0 an example code ran properly, so im assuming this env is fine documented some steps: https://t.co/0Tr0WYMGR2
0
0
3
Love the form factor of this thing: Day 1 > vscode tunnel > jupyterjub service > cloudflare tunnel to access ssh, jupyter hub over the internet.. Have dealt with aarch64 on a gh200s before while running rfdiffusion2, so hopefully the issues I encounter would be same 😅🤞🏻
2
0
8
Meet LLaMAX2! ✨ SOTA Translation: Excel in both high- and low-resource languages. ✨ Lightweight Pipeline: Built on a powerful Qwen3 instruct models, it requires parallel data only through smart layer-selective tuning. ✨ Elite Reasoning: Performance on par with Qwen3-Instruct.
3
4
5
Thrilled to announce our new preprint, “Protein Hunter: Exploiting Structure Hallucination within Diffusion for Protein Design,” in collaboration with Griffin, @GBhardwaj8 and @sokrypton 🧬Code and notebooks will be released by the end of this week. 🎧Golden- Kpop Demon Hunters
4
58
298
How to start a presentation? Most people start with all the boring info (e.g., repeating everything your host just said) and unmotivated technical details. 🥱 After a few mins, everyone starts checking their phone! 😱 How to avoid this? Here is a simple template. 🧵
66
1K
6K
Congrats to my colleagues @oanacarja's NSF CAREER https://t.co/z773AdvQWt and @YunWilliamYu's NIH MIRA https://t.co/iMfHbCUMmT 🎉their work exemplifies CMU comp bio: innovative & distinct methodology grounded in biological applications with real impact. @CMUCompBio @SCSatCMU
cs.cmu.edu
SCS faculty member Yun William Yu has received a Maximizing Investigators' Research Award (MIRA) from the National Institute of General Medical Sciences.
1
4
29
Left my engineering job and promotion to contribute to open source AI. Joined Aya project by @Cohere_Labs as one of the 3K volunteers and ended up publishing my first research paper as first author. Later joined the lab as a research engineer! Dream big. Take risks. Work hard :)
23
36
970
@nvidia Full details on how to retrain RF3 from scratch coming soon 👀
0
1
7
if you want to see how determined someone is.. ask them to compile dgl library from scratch for aarch64 😅
1
0
1