Aya Abdelsalam Ismail
@asalam_91
Followers
496
Following
380
Media
12
Statuses
101
Co-founder and CSO @guidelabsai Former Research Scientist @prescientdesign @Genentech, PhD @umdcs
Joined October 2019
RT @jiwoncpark: Come say hi at my #AISTATS2025 poster tomorrow!.📍 Poster Session I #83.🕒 Sat, May 3 at 3pm.Semiparametric Conformal Predict….
0
10
0
RT @amywang01: 📢 Excited to present our poster at the #ICLR2025 @gembioworkshop! I'll be introducing TherAbDesign - a novel, sequence-based….
0
19
0
2
7
96
I will be presenting this work tomorrow at #ICLR2025 at 10 am stop by to know how to build protein language models and use them to design proteins with new properties!.
[1/n] Does AlphaFold3 "know" biophysics and the physics of protein folding? Are protein language models (pLMs) learning coevolutionary patterns? You can try to guess the answer to these questions using mechanistic interpretability. But the thing is, more often than not, we know
0
10
43
I will be @iclr_conf this week! If you are there drop me a message I would love to catch up and maybe talk about interpretable foundation models 😉.
1
0
12
RT @nc_frey: Lab-in-the-loop therapeutic antibody design. At @PrescientDesign @genentech we have spent 3+ years reimagining drug discovery.….
0
123
0
RT @keunwoochoi: doing any data loading in Torch? i wrote a beginner's guide (as a beginner) to Torchdata 0.10.1, the first version with da….
0
13
0
RT @ml4proteins: Next Tuesday, 2/4 @ 4 pm EST, @asalam_91 and @nc_frey will present "Concept Bottleneck Language Models" .Paper: http….
arxiv.org
We introduce Concept Bottleneck Protein Language Models (CB-pLM), a generative masked language model with a layer where each neuron corresponds to an interpretable concept. Our architecture offers...
0
9
0
Exciting news! I've left Prescient to co-found @guidelabsai an interpretability startup with @juliusadml. We're building interpretable foundational models to address key AI challenges. While I loved my time at Prescient, I'm thrilled to build something I'm very passionate about!.
6
3
55
RT @tagasovska: Research Intern position @PrescientDesign in Switzerland 🇨🇭💊.join our Frontier Research team in 2025! .Apply now or share w….
0
7
0
More details in the paper: and our blog post You can also find the code on GitHub: and model weights on 🤗. You can also find.
github.com
Lbster: Language models for Biological Sequence Transformation and Evolutionary Representation - prescient-design/lobster
1
3
22
[4/n] Our architecture allows us to know what concepts the model learned and which concepts the model uses during inference by inspecting the weights of the final linear layer; this offers a way to debug and asses the model's quality.
2
1
10
[3/n] Interpretability: The concept bottleneck can be used to understand which concept the model uses to predict a certain amino acid. Reliably controlling model behavior: The concepts can be used as knobs to control the model's output.
1
1
10
[2/n] We train a mask language model with up to 3 Billion parameters with a layer that directly encodes biophysical and biochemical concepts that biologists care about. These models match the performance of unconstrained masked language model.
2
2
10
[1/n] Does AlphaFold3 "know" biophysics and the physics of protein folding? Are protein language models (pLMs) learning coevolutionary patterns? You can try to guess the answer to these questions using mechanistic interpretability. But the thing is, more often than not, we know
5
60
296
RT @tagasovska: Curious how causality can boost generative models in low-data regimes? ✨.Let's chat! Swing by our poster at #NeurIPS2024 to….
0
6
0