I'm incredibly proud of my former @UChicagoCS and @DSI_UChicago student Elena Orlova and collaborator Aleksei Ustimenko for their groundbreaking work building @simulacraai.
Forbes 30 Under 30 Europe 2026, Science & Healthcare. #ForbesUnder30
Building Simulacra AI so high-fidelity quantum chemistry simulation stops being a luxury.
Still early.
https://t.co/R6f4gKlLYh
Happy to invest in @simulacraai, who have just finished training the foundational model for quantum chemistry: the Large Wavefunction Model.
#NeurIPS2025
Appreciate the love @cyberfund 🫡
Reading-between-the-lines version: Every AI- biotech/materials company that raised $200M+ on the thesis “we’ll just run more DFT / MD” is now officially obsolete.
We solved the actual bottleneck.
2026 is going to be extremely painful for a lot of people.
The biggest bottleneck in drug and materials discovery is physics simulation. That bottleneck was just solved.
We're excited to invest in @simulacraai, which has completed pre-training the world’s first Large Wavefunction Model. It changes everything👇
https://t.co/VV4n0DK2hQ
The biggest bottleneck in drug and materials discovery is physics simulation. That bottleneck was just solved.
We're excited to invest in @simulacraai, which has completed pre-training the world’s first Large Wavefunction Model. It changes everything👇
https://t.co/VV4n0DK2hQ
We built this quietly.
Now it’s out: https://t.co/Jdz2IOH2H1
chemical accuracy is no longer rare or expensive.
Simulacra AI does chemically-accurate data generation 15–50× cheaper than Microsoft’s pipeline and beats CCSD scaling.
If you’re in pharma and want access → [email protected]
We’ve been quiet, building in semi-stealth.
The benchmark is now public: https://t.co/TaeZVfz6xe
Simulacra AI generates quantum-accurate data 15–50× cheaper than Microsoft’s Orbformer and >2–3× cheaper than CCSD, with CCSD(T)-level accuracy.
Ab-initio at scale is no longer impossible.
If you’re in pharma and want access → [email protected]
We’ve been quiet, building in semi-stealth.
The benchmark is now public: https://t.co/TaeZVfz6xe
Simulacra AI generates quantum-accurate data 15–50× cheaper than Microsoft’s Orbformer and >2–3× cheaper than CCSD, with CCSD(T)-level accuracy.
Ab-initio at scale is no longer impossible.
If you’re in pharma and want access → [email protected]
Происходящее с математиком Азатом Мифтаховым - пытка, растянутая во времени. По сфабрикованному делу ему дали шесть лет. За отказ давать ложные показания на невиновных Мифтахову создали условия поистине невыносимые. Его тюремный срок - шестилетний ад, и, когда этот ад закончился, Мифтахова арестовали на выходе из тюрьмы по показаниям лагерного стукача, сообщившего, что Мифтахов одобрил терроризм комментарием какого-то сюжета при просмотре телевизора.
Запредельная степень негодяйства.
Научное сообщество всего мира уже несколько лет добивается освобождения Мифтахова, я призываю их не ослаблять и не останавливать этих усилий. Я призываю всех требовать свободу Мифтахову и сделать так, чтобы история чудовищной подлости в его отношении была известна каждому.
Excited to announce that our work with @LProkhorenkova and @iseethefall on uncertainty estimation for tabular data via ensembles of gradient-boosted decision trees (GBDTs) has been accepted to ICLR2021!
Paper available here: https://t.co/trZfNdStWr
Introducing GraphGlove: represent words as graph nodes instead of vectors, get improved quality and learned hierarchy.
This work, done with Sergei Popov, Liudmila Prokhorenkova, and @lena_voita, will appear at #emnlp2020 — looking forward to discussions!
https://t.co/pDCTWzfYKB
Check out our new tutorial on uncertainty prediction in CatBoost and learn about different types of uncertainty – the one in data and knowledge uncertainty😺 https://t.co/g2ferhq4NV
Yandex ranks fifth in the list of companies with the highest publication count at ICML 2020! The ranking by Thundermark Capital is based on the Nature Index and takes into account the affiliation of each author. Find out more: https://t.co/d8qwy1I4DF
In our #ICML2020 paper we propose StochasticRank — the first globally converging learning-to-rank algorithm with provable guarantees that directly optimizes any ranking quality function. Already available in @CatBoostML!
▪️https://t.co/n2UgNCjuAz
[1/2] Happy to announce our work: Uncertainty in Gradient Boosting: https://t.co/7v6ioYLmQE
Gradient Boosted Decision Trees can sometimes be preferable to Deep Learning. Here, we use the ensemble-structure of GBDT models to estimate of total, data and knowledge uncertainty.