Seqera Co-Scientist Brings GPU Acceleration to Nextflow Pipelines with NVIDIA BioNeMo Agent Toolkit
The Future of Scientific Discovery is Agentic
Every process in life sciences is being redesigned with AI. Frontier models can now take on entire experimental workflows, from designing experiments to autonomous execution. Yet the fundamentals of science haven't changed. Science still depends on having a record of what was done, who did it, and with which data. More data demands better orchestration. More agents demand verifiable reproducibility. As we increasingly connect systems together, accessible APIs often defines software choices. When agents take on the work, automation and efficiency compound: faster, cheaper analysis buys more cycles, more iterations, more discovery. What has changed is that we now have the tools to apply these fundamentals autonomously, at a speed and scale that wasn't previously possible.
We've spent over a decade solving these problems for bioinformatics with Nextflow and Seqera. Now we're applying our learnings to agentic science as we build an intelligent engine for life sciences. Last month, we launched Seqera Co-Scientist, the collaborative agent that understands scientific context, reasons with you, and executes with intelligence. Co-Scientist is built to meet scientists wherever they work, including helping teams build, debug, optimize, and modernize their Nextflow pipelines with AI that understands their environment. Today, with the launch of the NVIDIA BioNeMo Agent Toolkit, we’re taking the next step. Teams can now more easily bring the power of accelerated computing and AI agents to their Nextflow pipelines.
Bringing NVIDIA BioNeMo Agent Toolkit into Co-Scientist
The NVIDIA BioNeMo Agent Toolkit is NVIDIA’s platform for enabling agentic workflows in life sciences. Bringing together NVIDIA Nemotron, NVIDIA NemoClaw, NVIDIA OpenShell, and NVIDIA BioNeMo, it provides the tools, models, and infrastructure for AI agents to advance biology across drug discovery, protein design, genomics, clinical data, and lab workflows. This includes specialized NVIDIA NIM microservices—NVIDIA-optimized containers to self-host GPU-accelerated inferencing microservices for pretrained and customized AI models— for life sciences, making it easy to deploy and run the right model at the right moment in any workflow.
BioNeMo skills are designed to work across the tools scientists already use, from querying biological datasets and preparing model-ready inputs, to launching reproducible workflows, analyzing outputs, and returning insights directly within the platforms teams depend on daily. As a first step, we're integrating the BioNeMo Agent Toolkit with Co-Scientist by leveraging Parabricks skills for GPU-acceleration of Nextflow pipelines.
NVIDIA’s Agent Skill for GPU-Acceleration of Nextflow
GPUs have become central to biological models. NVIDIA Parabricks, a scalable software suite providing GPU-accelerated versions of trusted, open-source tools, dramatically reduces runtimes for genomics pipelines and is regularly used in Nextflow workflows. But identifying where and how to integrate GPU acceleration into a Nextflow workflow has historically required deep expertise and benchmarking. NVIDIA’s Parabricks skills, part of BioNeMo Agent Toolkit, change this by guiding the integration of Parabricks into your existing workflows directly. Working as part of Co-Scientist (with full context across your pipelines, data, and execution history), this skill makes it straightforward to modernize and migrate to GPU-accelerated computing.
- →Inspect existing pipelines and identify processes that are candidates for GPU acceleration.
- →Suggest specific edits, guiding integration of Parabricks and other GPU-accelerated tools.
- →Generate a candidate revised pipeline with CPU-only applications optionally replaced by the equivalent Parabricks tool. At runtime simply provide the parameter to leverage the accelerated tools. This way you can run both and compare the performance of the pipelines.
- →Outputs include the revised pipeline and scripts to run both versions if a test configuration is found, as is the case for nf-core pipelines. The agent skill will also attempt to collect runtime metrics and generate a performance comparison report, giving teams a clear comparison of speed and efficiency gains before they commit.
Building the Foundation for Autonomous Research
At Seqera, we're continuing our partnerships on autonomous research: the ability for agents to autonomously explore the vast space of parameters, tools, and approaches to converge on better scientific outcomes. Co-Scientist and NVIDIA's BioNeMo Agent Toolkit accelerates this vision. Access to specialized NIMs and open models for drug discovery, protein design, and genomics simplifies how Co-Scientist's agents reach for the right specialized tool at the right moment in a research workflow, giving them more choices when selecting or tuning a model based on task, cost, and performance.
Building What Comes Next
Everything we've built is guided by a single conviction: science should be limited by ideas, not infrastructure. Agents that plan, execute, and optimize. Infrastructure that is reproducible, observable, and trustworthy at scale. Models and skills that understand the language of biology. Our collaboration with NVIDIA is a meaningful step in that direction. Together, we're building the foundations that will let agents work autonomously across the full scientific stack, with the enterprise-readiness and reproducibility that the world's leading life sciences organizations demand.