DDPS | Industrial Grade Scientific Machine Learning: Challenges and Opportunities by Santi Adavani
Description: There has been increasing interest in Scientific Machine Learning (SciML), which leverages advances in modern ...
Cofounder, Rocketml
Search every verified Santi Adavani interview, podcast appearance, and on-the-record quote — each transcript cross-checked by AI and human review to confirm speaker identity. Santi Adavani, co-founder and CTO of RocketML, has given multiple technical presentations and interviews between 2018 and 2022. In talks on scientific machine learning, he discussed the development of neural PDE solvers that use finite element losses to train networks without simulation data, and described algorithmic techniques such as multigrid training to reduce computational costs. He reported that a 256-cube training problem took around 2,000 minutes at a cloud cost of $735, while a multigrid approach completed the same task in 150 minutes for $55. He also noted that scaling out with more GPUs alone was too slow and unsustainable, and that RocketML built its own distributed data parallel engine after encountering convergence and performance issues with libraries like Horovod. In other appearances, Adavani offered career advice for aspiring data scientists, emphasizing fundamentals in statistics, linear algebra, and machine learning over certifications, and highlighted the importance of both technical and soft skills. He also led a hands-on tutorial on building computer vision applications from videos, covering object detection, pretrained models, and third-party APIs, and gave a talk on singular value decomposition (SVD), describing it as a versatile algorithm applicable across genomics, NLP, and other domains. He mentioned that RocketML computed SVD on a billion-by-million dataset with 400 singular values in about 15 minutes on 1,000 nodes.
Description: There has been increasing interest in Scientific Machine Learning (SciML), which leverages advances in modern ...
Technical Session D (Machine Learning): Scientific Machine Learning for Full-Field Predictions Using HPC Servers by Santi Adavani and Vinay Rao Thousands of PDE simulations at very high resolutions are necessary to solve design or inverse problems in several science and engineering applications. However, using traditional approaches, a single PDE simulation at these resolutions is computationally expensive with a long time-to-solution. We propose a solution to perform physics-based simulations at very high resolutions more than 50X faster than traditional finite-element and finite-difference…
Having the right education, training, and attitude is essential to become a Data Scientist. Santi Adavani, Founder and CTO @ RocketML, gives his perspective about what needs to be done to break into the Artificial Intelligence and Machine Learning industry. 🔴 Subscribe for more video like this: https://bit.ly/384DMaq 🔴 If you want to receive weekly updates, sign up for my newsletter: https://sendfox.com/rodrigowpa ------------- Connect with me ------------- Medium: https://bit.ly/3hQe2lu LinkedIn: https://bit.ly/3pHiWnw Photo by Joshua Sortino on Unsplash
Hands-on tutorials by Santi Adavani (CTO and co-founder of RocketML), to learn how to build amazing Computer Vision apps from videos within hours. You will learn computer vision, OpenCV, object detection, pretrained models from TensorFlow Hub, Google Vision API and RocketML tools. GitHub with tutorials - https://github.com/rmlhq/babyrocket/t... OpenCV - https://opencv.org/ TensorFlow Hub - https://tfhub.dev/ Google Vision API - https://cloud.google.com/vision RocketML - https://www.rocketml.net/
In June, 2018 -- RocketML CTO Santi Adavani gave a talk on Singular Value Decomposition at SparkAI Summit (https://databricks.com/session/from-g.... The title of the talk is “From Genomics to NLP, ONE ALGORITHM to RULE THEM ALL.” While, it may sound a bit exaggerated, SVD is indeed a very versatile algorithm/method for variety of applications ranging from -- yes, Genomics to NLP. Singular Value Decomposition (SVD) is a matrix decomposition technique with many applications in areas like genetics, natural language processing (NLP), and social network analysis. All these application areas result…
Sign in to search the full transcript archive, filter by topic, and access every quote from Santi Adavani.
The summary and quote tags on this profile are produced with AI assistance from verified, first-person interview transcripts, then checked by our team to confirm the speaker's identity and the accuracy of every quote. See how we verify →
Pro unlocks every transcript in full, plus unlimited search across every executive and company we cover.
50% off appliedPulling data at scale? The Data API gives you 30,000 transcripts a month over REST, MCP and JSON.
Compare all plans