This is Episode 1 of Structural Stories — a new podcast from Genia where we invite construction and engineering leaders to share what’s actually working (and what isn’t) as AI enters the field.
For our first episode, I sat down with Dr. Murat Melek, Senior Director of AI at Suffolk Construction. Murat is one of the most thoughtful people I’ve met at the intersection of structural engineering and artificial intelligence — someone who started his career doing hands-on structural design at Arup and Walter P Moore, then gradually moved toward visual programming, Python, and eventually full-on AI strategy at one of the country’s largest GCs.
We’ve been collaborating for almost two years now, and this conversation gave us a chance to reflect on what we’ve built together, where the industry is headed, and what young engineers entering the field today should make of all this.
A few things that stood out to me from our conversation:
On why automation makes sense for structural engineering Murat put it plainly: the structural workflow involves repeating the same process over and over. Building models, running loads, generating schedules — it consumes time without producing much learning. The question isn’t whether to automate, it’s where. His answer: framing and preliminary design are prime candidates because the guardrails are well-understood and the failure modes aren’t catastrophic. The critical judgment calls — shear checks, deflection, coordination — still need an engineer in the loop.
On how Suffolk actually uses Genia’s API Rather than putting structural engineering software in front of architects or pre-con staff who don’t want to use it, Suffolk’s team preps and transforms building model data into a schema Genia can process, then pipes the results back into their workflow. The use case that kicked it off: a series of one-story steel buildings across multiple locations with varying geometries. Being able to change a parameter, hit send, and get a structural response back in seconds is genuinely valuable when an architect is finalizing a layout under deadline.
On trust and accuracy This is where things get nuanced. Murat made a point I think about a lot: engineers today spend so much time building the model that they have almost no time left to interrogate the results. AI should flip that ratio. You want high-level benchmarking and element-level detail available simultaneously — not as a replacement for engineering judgment, but as a way to make that judgment faster and better-informed.
On the shift to outcome-based services We spent some time on a question that I think will define the next decade of AEC: does the client care how many hours went into a structural design, or do they care that it’s accurate, well-coordinated, cost-efficient, and doesn’t generate RFIs on site? Murat’s view: the technology is now here to make outcome-based billing inevitable. Licensure and stamping aren’t going away — accountability still matters — but the billable-hour model for repetitive design work is on borrowed time.
On the workforce question This is the one I keep coming back to. If AI is handling the work that used to train junior engineers, how does the next generation develop intuition? Murat’s take is that this is ultimately a user experience problem: we need to design systems that give young engineers feedback on the consequences of design decisions, not just output that makes decisions invisible. Domain expertise, planning, and the ability to translate engineering intent into language AI can act on — those are the skills that will matter most.
You can also watch or listen to the full episode on YouTube. If you have thoughts, questions for future guests, or someone you think we should have on — I’d love to hear from you.
— Zhihao
