Suggestions
Danny Reidenbach
AI Research Scientist | Generative AI for Drug Discovery | EECS & ChemE @ UC Berkeley
Danny Reidenbach, a senior at the University of California, Berkeley, is a talented and driven student with a dual degree in Chemical Engineering and Computer Science. He is particularly interested in exploring the applications of Deep Learning in both synthetic and organic systems, and has hands-on experience in software engineering, computational biology research, and machine learning.
Throughout his career, Danny has worked with numerous prestigious organizations including NVIDIA, Canny Lab, Yosef Lab, Google, Frontier Bio, Centrillion Biosciences, Inc., Bay Area Neutron Group, and Taco Bell. As an NLP Research Scientist at Canny Lab, he demonstrated his expertise in natural language processing by working on groundbreaking algorithms to improve and optimize chatbots. As a Machine Learning Research Intern at NVIDIA, Danny developed deep learning models to optimize image super-resolution for medical applications.
Danny is well-known for his most recent NLP project, which uses BERT architecture and transformers to predict Yelp ratings from reviews. In addition, Danny is an accomplished leader and has taken on many roles as a teaching assistant, camp counselor, and team assistant in the past.
Danny's background is impressive and shows a passion for pursuing new technologies and the exploration of their potentials. He would make an excellent addition to any team focused on synthetic and organic systems development.
Highlights
Excited to see my recent work with MolMIM highlighted by Jensen at the GTC keynote yesterday! For more details about how we are using gradient-free structure-based guidance for state-of-the-art molecule generation, stop by my posters at the GEM and MLGenX workshops at ICLR in Vienna in a couple of weeks (public paper coming soon).
Transforming this research endeavor into a scalable product took a massive team effort. Special thanks to everyone in the BioNeMo team and greater NVIDIA who were involved in building out MolMIM, UniDock, and the NIM infrastructure. Moving beyond academic benchmarks is crucial for further drug discovery development and I am happy to be part of a diverse and skilled team to make this possible. We have a lot more planned and are actively hiring expert scientists, engineers, and product managers to help deliver on our goal of accelerating meaningful drug discovery. For more info about what we do check out https://lnkd.in/gqs26KNK.
If you are curious about what makes MolMIM work, check out our paper below. Also please reach out if you have any questions!
MolMIM: https://lnkd.in/gRpf3k3E EvoSBDD(MolMIM + UniDock) public paper coming soon: check it out at ICLR MolMIM Playground: https://lnkd.in/geA3vcaH
Introducing CoarsenConf, an SE(3)-equivariant autoregressive generative model for molecular conformer generation. CoarsenConf is a hierarchical variational autoencoder that uses a variable length coarsening procedure to generate robust (i.e. consistently accurate) low-energy 3D coordinates for small molecules.
CoarsenConf essentially breaks down a molecule into intuitive subgraphs and generates the low-energy coordinates one subgraph at a time, filling the 3D volume from the inside out. CoarsenConf further bridges the gap between recent advances in geometric deep learning and ML-accelerated physics-based molecular simulations.
In addition to prior benchmarks, we also introduce new metrics that better measure robust generation. These metrics focus more on what scientists want these kinds of models to be able to do outside of an academic setting. We show that CoarsenConf specializes in having the lowest average and worst-case error across all tested molecules.
This work would not have been possible without the help and support of my advisor Aditi Krishnapriyan and her lab at UC Berkeley. If you are interested in learning more about the specifics or generative models for life sciences in general, please check out our preprint and blog post!
blog: https://lnkd.in/ggC8bVPA paper: https://lnkd.in/gh4Fus9Z code: coming soon
#ai #research #machinelearning #computationalbiology #computationalchemistry #drugdiscovery #ml #deeplearning
