Muhammad Saad Naeem

Somewhere between engineering, computational chemistry & ML 🧑‍💻

[Muhammad Saad Naeem @saadnaeem_]

When Molecular Dynamics Took Too Long https://t.co/doJeVni0Zu

[TheorHetCat ICIQ @TheorHetCatICIQ]

Today is a great day! Saad will be defending his thesis. The last student of the @solar2chem team, presents an amazing research trip and our first in collaboration with Ricard Garcia-Valls @universitatURV. With one of the first uses of MLIPs for transport in membranes.

[Muhammad Saad Naeem @saadnaeem_]

I revisit the classical HER beyond low-coverage Sabatier
constraints and found a alternative HER pathway and its theoretical descriptors. Metal centers pre-concentrate H and form di-hydrogen complexes at monolayer coverage, explaining the high experimental activity.

[TheorHetCat ICIQ @TheorHetCatICIQ]

Excellent news coming... trying different synthesis routes is always fun. Thanks to Menny for the leadership and Saad for smoothly running our simulations!

[The Nobel Prize @NobelPrize]

Is this one of the most famous equations in physics?

The Schrödinger equation is a cornerstone of quantum physics - the analogue of Newton's second law for quantum mechanics. Its derivation led to Schrödinger, born on this day in 1887, receiving the 1933 Nobel Prize in Physics.

[J. Am. Chem. Soc. @J_A_C_S]

On the cover of this week's issue: "Light-Driven Chemical Cascade Reduces Barriers to Hydrogen Production"

Read it here 🔗 https://t.co/vFgBojIRpd

[Muhammad Saad Naeem @saadnaeem_]

We introduce a two-step, light-driven cascade for photocatalytic H₂ production, eliminating noble-metal costs and expanding semiconductor choices.

[TheorHetCat ICIQ @TheorHetCatICIQ]

So happy that this collaboration with @at_bgu Menny's group went so nicely.
We love cascades! Enjoy the Solstice and thanks Saad for your great computational work!
https://t.co/gmiemuuMR3

[Chris Piech @chrispiech]

The GPT Surprise😮! We ran one of the first controlled experiments on the impacts of ChatGPT for learning. We ran it in a massive #python course @StanfordEng. Many classes have deployed AI tutors, but very few have studied the causal effects on a #education. Can you guess the impact?... 🔬🎓

[Aditi Krishnapriyan @ask1729]

1/ Machine learning force fields are hot right now 🔥: models are getting bigger + being trained on more data. But how do we balance size, speed, and specificity? We introduce a method for doing model distillation on large-scale MLFFs into fast, specialized MLFFs!

[Muhammad Saad Naeem @saadnaeem_]

Stuck at my fraud detection project. Below is the best result using XGBoost and after tuning class weights on a highly imbalanced dataset (600:1). Selective sampling of any kind or F.Eng produces inferior results. Targetting high recall. Any suggestions?

[Grant Sanderson @3blue1brown]

I just posted a new video, originally made as part of a new exhibit at the Computer History Museum on the history of chatbots.

My hope is that it offers a lighter-weight, but still substantive, intro to Large Language Models, complementing the more technical breakdown of the underlying architecture that other videos of the channel cover.

https://t.co/xZBIq3S9QM

[minami @minamisatokun]

I use Python cause I'm a chill guy not because I have skill issues

[Muhammad Saad Naeem @saadnaeem_]

@svpino New to ML and doing baby projects like price prediction and fraud detection. Should I bother deploying them? I feel like I am learning but seeing the current rate of progress, I thinm I should be doing RAG or something related to LLMs.

[Grant Sanderson @3blue1brown]

Happy Tau Day! As something a little different, here’s a clip from a bit I did at @standupmaths's “Evening of Unnecessary Detail” a few years ago.

Full video here: https://t.co/BNlCsRaCQY

[Muhammad Saad Naeem @saadnaeem_]

@Fidias0 1- Create "Productive" jobs. Regulation and admin jobs must decrease. 2- ⬆️ R&D budget +add startup investments. Fund innovative risk-takers and local manufacturing and technology.

[Muhammad Saad Naeem @saadnaeem_]

I am confused. After outlier removal, manual feature eng and insistence on using simple models (LinReg, Ridge), I got the following results for my house price prediction. Should I shift to DL models for better results or explore feature engineering further? 🤔 #newbie

[Muhammad Saad Naeem @saadnaeem_]

Laying my ML foundation by completing Mathematics for Machine Learning and Data Science by https://t.co/lT4yJ2rska. Learned and coded PCA, Gradient Descent and a complete Neural Network from scratch. https://t.co/VnGU3LCiSc #Coursera

[Muhammad Saad Naeem @saadnaeem_]

@MushtaqBilalPhD What does an academic family even mean? A players nuture A players. One can argue about the scale of funding in such labs but that doesn’t promise a nobel.

[Muhammad Saad Naeem @saadnaeem_]

Picture Credits:3Blue1Brown

[Muhammad Saad Naeem @saadnaeem_]

We all memorized that the derivative of x² is 2x, but do we actually know why? Let’s actually see it visually—I promise it’s more satisfying than just memorizing!
🧵

[Muhammad Saad Naeem @saadnaeem_]

After rearranging, we finally get: