AI is becoming an increasingly important tool for drug discovery. However, for many pharmaceutical and biotech companies, the challenge isn’t finding more powerful models. It’s getting existing models to work together with proprietary research data and computing infrastructure in a way that’s actually useful for scientists.
That is exactly the problem Databricks and Nvidia are trying to address with the new Genesis Workbench – an open-source blueprint for building AI applications in the life sciences.
Rather than introducing a new model, the project combines enterprise data, Nvidia’s BioNeMo models and GPU infrastructure into a single environment designed to help researchers move from setting up AI workflows to using them.
In a Databricks blog post, the company described Genesis Workbench as “an open blueprint for a life-sciences application on Databricks – a modular workbench that brings the major stages of computational drug discovery under one roof, one UI, and one governance model.”
Drug discovery projects often combine multiple AI models with different components that all play a role in data-driven scientific research. These components may include internal research data, lab results and GPU resources. The problem is that these components frequently live in separate environments. This makes collaboration and reproducibility more difficult than they need to be.
The number of AI models available to researchers is only growing. That makes it even more important to have a way to manage them alongside proprietary data and existing research workflows.
(Credits:Databricks)
Genesis Workbench is Databricks’ attempt to change that. Instead of focusing on one part of the drug discovery process, it pulls together tools for genomics, single-cell analysis, protein engineering and small molecule design into a single environment where researchers can move between tasks more easily.
According to Databricks, “By centralizing both public and proprietary datasets with Databricks AI Search, we’ve entirely eliminated external API dependencies. Ultimately, this seamless setup connects every step of the process—allowing genomics findings to flow effortlessly into single-cell validation, target structure prediction, candidate docking, ADMET, and ranking.”
Databricks says the platform relies on open-source models managed through Unity Catalog, with MLflow tracking experiments and GPU-backed Model Serving handling inference.
Nvidia contributes its BioNeMo Agent Toolkit along with technologies such as Parabricks and a growing portfolio of biology and chemistry models that can be incorporated into scientific workflows.
One of the platform’s distinguishing features is that it runs entirely inside a customer’s Databricks environment. That is important, because it allows organizations to keep sensitive research data within existing governance controls instead of sending it to third-party AI services.
There is also flexibility to scale as biological AI continues to evolve. New AI models will continue to emerge. Organizations can add or replace individual modules without rebuilding the broader research environment.
Genesis Workbench also reflects a broader shift in how enterprise AI platforms are evolving. Much of the industry’s early focus centered on building larger and more capable foundation models. However, that is changing. At BigDatawire, we see vendors are increasingly focusing (and competing) on how well those models can be integrated with enterprise data and domain-specific workflows.
(DCStudio/Shutterstock)
Life sciences, including drug discovery research, present a particularly demanding environment as research spans multiple disciplines. This includes everything from genomics and structural biology to chemistry and clinical data, while also having to handle highly regulated and proprietary data.
Building AI applications in this setting requires more than raw computing power. It also demands secure access to data and the flexibility to incorporate new models as the technology evolves.
For Databricks, Genesis Workbench is another example of the company pushing beyond analytics and further into AI applications built on the lakehouse. Nvidia, meanwhile, is using the project to put BioNeMo and its accelerated computing software at the center of enterprise drug discovery workflows instead of treating them as standalone research tools.
If researchers spend less time moving data, connecting models and configuring infrastructure, they have more time to focus on the science. That is what Nvidia and Databricks are trying to achieve with the Genesis Workbench.
The post Databricks and Nvidia Launch Genesis Workbench for AI Scientists appeared first on AIwire.