Ahmed Elnaggar

AI and Deep Learning Specialist

[Ahmed Elnaggar @Elnaggar_AI]

🌟 Introducing Ankh V3 ☥ 🌟

Our optimized protein language model (PLM) is setting new benchmarks in the field! Achieving state-of-the-art performance with self-supervised only.

Dive into the details: https://t.co/nHJfmuBlHI

(1/n)

#ProteinModel #AI #Innovation #Bioinformatics

[Ahmed Elnaggar @Elnaggar_AI]

@Google @googlecloud 🌍 A Nod to Google Support 💚

We appreciate Shira, Jonathan, and their teams for their unwavering support throughout our journey with Ankh. Your assistance made all the difference!

#Google #googlecloud

(5/n)

[Ahmed Elnaggar @Elnaggar_AI]

🌟 Introducing Ankh V3 ☥ 🌟

Our optimized protein language model (PLM) is setting new benchmarks in the field! Achieving state-of-the-art performance with self-supervised only.

Dive into the details: https://t.co/nHJfmuBlHI

(1/n)

#ProteinModel #AI #Innovation #Bioinformatics

[Ahmed Elnaggar @Elnaggar_AI]

@Google @googlecloud ⚙️ Precision Engineering 🛠️

Guided meticulously engineered training experiments led to optimizing our transformer PLM for protein modeling — every detail from masking strategies to multitask learning was fine-tuned while keeping parameter counts low! 📉

(4/n)

[Ahmed Elnaggar @Elnaggar_AI]

🚀 Google just dropped Gemma 3 – the most powerful open multimodal LLM yet!

���� 4 model sizes (1B, 4B, 12B, 27B)
🖼️ Text + Vision (4B and up)
🌍 140+ languages
📈 128K context

Crushing benchmarks, pushing boundaries. Ready to try?
👉 https://t.co/P77AsHYWns
#AI #LLM #google

[Ahmed Elnaggar @Elnaggar_AI]

🚀 Day 1 at #GDE AI Train-the-Trainer Bootcamp in Paris was nothing short of amazing!

🔥 Deep dives into #Gemini & #Gemma
💡 Expert-led workshops & keynotes
🤝 Connecting with the best in AI & cloud

Can’t wait for what’s next! Who’s here with us? 👀 #BuildWithAI #GoogleCloud

[Ahmed Elnaggar @Elnaggar_AI]

@_akhaliq Now it is fully integrated into @huggingface library :)

https://t.co/KkxNNC7Nts

https://t.co/JViVQleETl

https://t.co/7voCIWs13g

[Ahmed Elnaggar @Elnaggar_AI]

If you have limited memory, you could use the base or the large versions.

Base Version:
https://t.co/KkxNNC7Nts

Large Version:
https://t.co/JViVQleETl

[Ahmed Elnaggar @Elnaggar_AI]

Are you working with long multi-language documents and looking for an efficient large-language model?

You can now use the efficient MLongT5 model that supports 102 languages from the @huggingface library.

XL Version:
https://t.co/7voCIWs13g

Paper:
https://t.co/6WpApyNxZV

[Ahmed Elnaggar @Elnaggar_AI]

I extend @elonmusk concerns to other LLM domains like biochemistry and bioinformatics.

I believe everyone in such critical fields should work together and think about how we could avoid such dramatic outcomes. (n/n)

[Ahmed Elnaggar @Elnaggar_AI]

@elonmusk announced an initiative to stop working on LLM like GPT4 for 6 months due to increase concerns.

Ok, but should we focus our concerns on a more dangerous type of LLM in bioinformatics and biochemistry? (1/n)

[Ahmed Elnaggar @Elnaggar_AI]

As I work in this field, I believe there will be an LLM in the next 2 years that will allow anyone to create computationally at low resources a new unique drug or virus?

I think this thought should be more terrifying than anything else. (2/n)

[Ahmed Elnaggar @Elnaggar_AI]

@YiTayML When the code and checkpoints will be available ?

[Ahmed Elnaggar @Elnaggar_AI]

@NicoLopezCarran Thanks @NicoLopezCarran :)
One small step in the right direction and more to come.

[Ahmed Elnaggar @Elnaggar_AI]

Introducing Ankh ☥ :
An optimized protein language model (PLM) achieving SoTA performance with < 10% of the top/closest-performing models’ parameters. It is benchmarked on various downstream protein tasks, and tested for protein generation.
https://t.co/7VYwz748TV
(1/n)

[Ahmed Elnaggar @Elnaggar_AI]

@hazemessamm It was my pleasure to work with such a talented, educated, and hardworking person like you.
I believe you are a new star coming to the deep-learning world and will make a big impact :)

[Ahmed Elnaggar @Elnaggar_AI]

@Google @googlecloud @huggingface @proteinea @hazemessamm @wafaaashraf97 @abs_WaleedM @KerdawyBioAI @Columbia @MoAlQuraishi @charochereau @TU_Muenchen @rostlab @BurkhardRost @julien_c @ClementDelangue @colinraffel @ada_rob @fair @MetaAI @FacebookAI @joshim5 @alexrives @ylecun @nvidia @NVIDIAAI @MSFTResearch @KevinKaichuang Finally, you can read more about the ANKH model and its application in the industry on the @proteinea blog: https://t.co/xZJ5DC8H5f

For commercial licenses/collaborations, contact @proteinea .

For academic support/collaborations, contact either @rostlab or @proteinea .
(11/n)

[Ahmed Elnaggar @Elnaggar_AI]

@Google @googlecloud @huggingface @proteinea @hazemessamm @wafaaashraf97 @abs_WaleedM @KerdawyBioAI @Columbia @MoAlQuraishi @charochereau @TU_Muenchen @rostlab @BurkhardRost @julien_c @ClementDelangue @colinraffel @ada_rob I want to personally thank:
@FAIR @MetaAI @facebookai : @joshim5 @alexrives @ylecun
@nvidia : @NVIDIAAI
@msftresearch : @KevinKaichuang
for their great contribution to the bioinformatics field, especially the PLMs direction.
All our work has changed this field forever.
(10/n)