Has agentic orchestration replaced the workflow engine?
Agents change who decides what happens next. They do not remove the need to execute those decisions reliably, safely and with scientific provenance.
Read articleScience Architects · By Adrian Rossall
The architecture
behind scientific AI.
Independent thinking on AI architecture for science. I write about how agents, data, compute and scientific workflows fit together, drawing on two decades of building scientific platforms.
Read the articles
Practical thinking for leaders building AI-enabled scientific platforms.
04
Scientific intent
Hypotheses · decisions · accountable outcomes
03
Operating model
People · workflow · governance
02
Scientific ecosystem
Data · models · instruments · agents
01
Platform foundations
Identity · provenance · compute · observability
Scientific intent sets the direction. Foundations make the evidence trustworthy.
01 / WRITING
Scientific AI, viewed through an architect’s lens
Notes on the architecture beneath modern scientific discovery: where data, computation, laboratories and AI agents become one system.
Three frontier labs moved into drug discovery. None of them built a platform.
Frontier models are rapidly commoditising scientific reasoning and tool use. The durable value is moving to project state, provenance and the joins between systems.
Read articleThe 3rd generation of AI drug design platforms is built on agents and context engineering.
How platform foundations, scientific ecosystems and agentic science build on one another.
Read article02 / ARCHITECTURE
Scientific platforms need a coherent system of evidence.
The architecture has to connect scientific intent all the way down to the foundations that make a result reproducible, explainable and useful.
Explore a scientific AI platform.
Follow the connections between agents, workflows, scientific tools and evidence.
Start with the decision
A scientific platform is only useful when it improves the quality, speed or confidence of a decision. Understand that decision before choosing technologies.
Engineer the joins
Value is created between data, models, instruments, workflows and people. Those boundaries and contracts deserve more attention than any individual tool.
Make provenance structural
Identity, lineage, reproducibility and accountable human judgement are platform capabilities. They cannot be retrofitted as reporting features.
Design the organisation too
Systems inherit the boundaries of the teams that build and operate them. Architecture must account for ownership, incentives and how teams manage change.
03 / ADRIAN ROSSALL
Scientist by training. Platform architect by practice.
I work at the point where scientific ambition becomes an operable system.
Over two decades, I have designed and led scientific platforms spanning laboratory automation, research data, computational chemistry, machine learning, scalable compute and AI-enabled drug discovery. That experience informs the architecture and writing I share here.
Full profile on LinkedInSCIENCE ARCHITECTS
Keep exploring scientific AI.
Follow my writing and join the discussion on the architecture behind scientific discovery.
Connect on LinkedIn