Computational Design in Practice, ft. Perkins&Will

In this hour-long conversation, Stephen Drew is joined by Mario Romero, a digital practice and computational design lead at Perkins&Will, and Nirmala Srinivasa, a project manager at the firm. Together they walk through a real tech-campus project and show how parametric and computational methods move a design from sketch to signed-off, built reality.

Who this is for

Architectural designers, technologists, project managers and practice leaders who want a grounded, jargon-free view of how computational design actually fits into day-to-day project delivery. No coding background is assumed. It will also suit anyone weighing up how parametric tools and AI sit alongside Revit and established BIM workflows.

Learning outcomes

By the end of this lesson you will be able to:

1. Explain what computational design means in a practice setting, and why it is closer to driving design with parameters and data than to advanced programming.
2. Describe how real-time optioneering helps secure client sign-off faster while keeping the design vision intact.
3. Identify where parametric workflows reduce rework compared with constraint-led modelling.
4. Recognise how design data can be reused to bring contractors and fabricators on board and rationalise an assembly for buildability.
5. Discuss the role of data interoperability across tools such as Rhino, Grasshopper, SketchUp and Revit.
6. Form a considered view on where AI may support, rather than replace, architectural work.

Local studios, one world lab (04:15)

Nirmala and Mario describe Perkins&Will as a global practice of more than 2,800 people, with studios across the United States, the United Kingdom and Northern Europe and a stated ambition to grow in Asia. The firm frames itself as a network of local studios backed by a shared pool of expertise, with centres of excellence in areas such as healthcare and corporate work that teams can draw on when a design challenge exceeds a single studio's experience.

From SketchUp sketch to white model (09:41)

The project at the heart of the episode is a two-phase tech campus, started in early 2018. The pair focus on a top-floor collaborative space the client called the lodge, featuring a set of curved fins above a connecting stair. Early SketchUp studies were single, static one-shots: useful for a quick look, but discarded the moment a client preferred a different direction.

Why iteration in Revit became the bottleneck (13:48)

Moving the fins into Revit produced a white model the client liked, but each variation, making fins thicker, taller, more or less spaced, or changing the curve, meant rebuilding geometry by hand. Mario and Nirmala describe the cost of that: time lost to remodelling and archiving, and the risk of stalling client momentum when iterations cannot keep pace with a conversation.

Computational design, defined simply (19:48 and 31:53)

Mario reframes computational design as using parameters and algorithms to drive design constraints, nothing more exotic than that. In his words, you can be a computational designer with a spreadsheet linked to another tool. For the lodge, a Grasshopper model in Rhino let the team flex the fin profiles freely, sketching within a design space rather than fighting the hosting rules of a constraint-led platform.

Using the same data to win over the contractor (29:00)

A key turning point: once the client chose a version, the same parametric data was disassembled to show the general contractor and subcontractor how the fins broke down into standard and bespoke parts, almost like a kit of components. Translated into flat CAD the fabricators could read, a feature that looked complex became feasible, and the contractor became an enthusiastic collaborator rather than a blocker.

Data interoperability and being data agnostic (33:17)

Mario describes the firm's aim to be data agnostic: workflows and scripts are designed to be reusable across project types and studios, not one-offs tied to a single job. Because clients and consultants often work in Revit, delivery still lands there, so the computational designer's role is partly that of a data translator, making sure Rhino, Grasshopper, SketchUp and Revit all talk to one another.

The craft behind seamless design (36:24)

Nirmala makes the case that great design often looks simple precisely because the effort behind it is hidden. The clean white model sat on top of a dense, complex script. Both guests return to a theme: hold the integrity of the design concept first, then find the partner, whether a colleague, a subcontractor or a tool, that lets you realise it.

Where AI fits (46:34)

Looking ahead, Mario is positive but measured. He sees near-term value in AI handling low-value tasks such as meeting minutes and rapid visualisation from a viewport screenshot, always on a trust-but-verify basis, freeing designers for design. Longer term, he points to the long-standing promise of the model as deliverable, noting that BIM has been around for roughly three decades, and suggests AI may help architecture finally deliver richer models rather than flattened documents.

Key terms

• Computational design: using parameters and algorithms to drive design decisions and constraints.
• Parametric model: a model whose elements are controlled by adjustable inputs, so changes ripple through automatically.
• Optioneering: rapidly generating and comparing design options, here in real time with the client.
• Grasshopper and Rhino: a NURBS-based modelling environment and its visual scripting tool, used for flexible, sketch-like iteration.
• Data interoperability: keeping design data usable across different software so tools can exchange information.
• BIM: building information modelling, the data-rich approach to designing and documenting buildings.

Reflective prompts for your CPD record

1. Where in your current projects does rework slow down client sign-off, and could a parametric approach shorten that loop?
2. How might you reuse design data to bring contractors and fabricators into a conversation earlier?
3. Which low-value tasks in your week could be delegated to AI on a trust-but-verify basis, and what would you do with the time saved?

About the guests

Mario Romero is a digital practice and computational design lead at Perkins&Will, with around 17 years of experience that began at a supertall skyscraper design firm. Nirmala Srinivasa is a project manager at Perkins&Will, based in the Dallas studio, who has worked across corporate interiors, aviation, healthcare and education since joining the firm in 2017. Perkins&Will is a global architecture and design practice. Watch the full conversation above or listen on Spotify.