4/4 We went from "Can AI understand molecules?" to "What should AI design next?". Complete technical deep-dive:, .

3/4 The breakthrough: Generation revolution ⚡. Regression Transformer showed bidirectional conditioning works. Chemformer (BART transformers) + MCTS for synthesis planning.

2/4 The evolution: • ChemBERTa vs MolBERT: Scale vs efficiency battle • SMI-TED: Learnable compression creating representations that self-organize by chemical principles • AI discovering chemistry without being explicitly taught

1/4 Can transformers read molecules like text?. Turns out CCO (ethanol), c1ccccc1 (benzene) are just sequences waiting to be tokenized. But chemistry posed unique challenges—from 42 to 52K token vocabularies needed.

Thread: How Transformers Learned Chemistry (3-Part Blog Series). I wrote an intro of chemical AI for computer scientists - from molecules as text to AI designing new drugs. Here's the journey: 🧵.

I found it difficult for data scientists to enter protein language models. So, here is my attempt to reduce the gap:

Tracking the progress of transformer all over the board is my weekend reading theme these days. Have created some of those including the one this week. This one is about Self Supervised Learning in Vision with Regression: MAE, SimMIM, and MaskFeat.

Another long weekend, another blog. This time is a type of blog that I want to write since some times: a back to basic blog: Knuth-Morris-Pratt and Aho Corasick algorithms.

My weekend blogging . This time is about intro to codon optimist

Every programmer should embrace code assisted AI. If you are still not convinced, I record my experience working with Amazon Q Developer in my blog here: .Enjoy.

My third blog of the year: @JonatanSelsing.

My 2nd blog published in official AWS blog. This time is about Uber H3 UDF in Athena: .

RT @DGLGraph: DGL 1.0 has arrived! Huge milestone of the past 3+ years of development. 👉Check out the blog for the release summary and the….

RT @JonatanSelsing: We wrote a bit on how we approach application design when thinking about data management at scale! Give it a read, and….

RT @MalwareJake: Security vendors demonstrating how their products "would stop the Uber breach."

RT @AmazonScience: Tens of thousands of customers use Amazon Redshift in the AWS global infrastructure to process exabytes of data daily. A….

RT @clrs4e: Pulitzer Prizes were just announced, and Introduction to Algorithms has maintained its 32-year streak of not winning one.

Got my copy of Introduction to Algorithms 4th edition book today.

Siap jadi supporter Hendra Setiawan & Fajar Alfian #frenchopen2021