“What if detecting critical structural defects was as easy as uploading a photo?”

Most structural inspections still rely on manual site visits, ladders, and visual checks, leaving room for missed defects, costly delays, and safety risks. In this episode I sit down with Ralph Noronha, founder of Defectech, to explore how AI, drones, and 3D modeling are transforming the way engineers detect and manage structural defects. Ralph shares the story behind building a bootstrap startup that turns ordinary images into actionable insights, enabling faster, safer, and more scalable inspections without sacrificing accuracy.

From the technical challenges of training bespoke machine learning models to the human side of entrepreneurship, Ralph explains how Defectech balances risk, innovation, and real-world impact. Whether you’re an engineer, asset owner, or tech enthusiast, this conversation reveals how modern tools are redefining inspection workflows, making defect detection as easy as uploading a photo, and why the future of structural intelligence is about combining technology with practical expertise.

🔑 Key Takeaways:

- Traditional inspections miss defects; AI and drones make detection faster, safer, and more complete.

- Photos become actionable 3D models, turning raw data into precise insight.

- Bespoke machine learning models require meticulous data, annotation, and hyperparameter tuning.

- Effective adoption depends on human judgment, trust, and workflow design not just technology.

- Founders succeed by leveraging strengths, iterating fast, and learning from failure under tight resources.

⏱ Timestamps:

00:00 - Engineering Meets Innovation

04:53 - From Banking To Startup

08:56 - Founder Mindset and Motivation

12:23 - AI-Powered Defect Detection

17:53 - Building and Training AI Models

22:05 - Using Defectech In Practice

26:48 - Recognition, Adoption, And Future Insights

31:27 - Asset Storytelling And Predictive Insights

35:22 - Customer Focus And Common Defects

40:53 - AI Impact And Closing Thoughts

🔗 Guest: Ralph Noronha

Co-Founder of Defectech

Website: https://www.defectech.com.au/

LinkedIn (Company): https://au.linkedin.com/company/defectech

LinkedIn (Ralph Noronha): https://au.linkedin.com/in/ralnor

🎧 Host & Links

LinkedIn: https://www.linkedin.com/in/owenrkeenan

Spotify: https://open.spotify.com/show/7vcBFpl87NcqpLf0seIRaW

YouTube: https://www.youtube.com/channel/UCE9Ax8jAJl2x0QQp3cO8vrw

VoiceInspect.ai (AI for inspections): https://www.voiceinspect.ai

👉 Subscribe to the Structural Intelligence Podcast on YouTube and Spotify to stay updated on new episodes.

#Engineering #Technology #Innovation

“What if the biggest risk to our infrastructure isn’t corrosion, cracking, or ageing, but the assumptions we still make?”

Most asset inspections rely on “representative sampling”, checking a handful of areas and extrapolating the rest. But in complex, ageing buildings, that approach can quietly hide defects, increase risk, and lead to expensive surprises later.

In this episode of the Structural Intelligence Podcast, host Owen Keenan sits down with Derek Feebrey, Co-Founder and CEO of Trendspek, to explore why sampling often falls short, and what a better future looks like, one built on complete, auditable data.

From aviation-inspired precision to millimetre-accurate 3D digital records, Derek explains how the industry is shifting from static PDF reports to living models of the built environment, and why the biggest barrier is not technology, it’s trust, process, and change management.

If you work in engineering, asset management, facilities, or infrastructure, this episode will challenge the way you think about inspections, evidence, and decision-making.

🔑 Key Takeaways:

• Why “representative” inspections can miss critical risks
• How complete data reduces assumptions and improves decisions
• Why 3D modelling is changing how we plan remediation and budgets
• The role of engineers, contractors, and owners in better collaboration
• Why adoption is more about people and process than technology

⏱ Timestamps:00:00 - Why Sampling Fails06:02 - From Photos to 3D Insight09:01 - Digital Inspections from Data13:42 - Agnostic Data Capture17:31 - The Limits of Representative Inspections22:40 - Remediation and Cost Certainty26:32 - Optimizing Asset Management30:19 - Working Together Virtually35:29 - Digitizing Data and Building Smarter Insights38:26 - AI-Powered Infrastructure

🔗 Guest: Derek Feebrey

Co-Founder & CEO TrendspekWebsite: https://trendspek.com/LinkedIn (Company): https://linkedin.com/company/trendspekLinkedIn (Derek Feebrey): https://linkedin.com/in/derek-feebrey

🎧 Host & Links

LinkedIn: https://linkedin.com/in/owenrkeenanSpotify: https://open.spotify.com/show/7vcBFpl87NcqpLf0seIRaWYouTube: https://www.youtube.com/@Structural-Intelligence

VoiceInspect.ai (AI for inspections): https://www.voiceinspect.ai

👉 Subscribe to the Structural Intelligence Podcast on YouTube and Spotify to stay updated on new episodes.

“When engineers start trusting AI, what exactly are they trusting?”

In this episode of The Structural Intelligence Podcast, host Owen Keenan sits down with Stijn Jansen, co-founder of VIKTOR, to explore what happens when engineers begin building tools that build themselves. From aerospace engineering to software automation, Stijn explains how VIKTOR enables engineers to replace brittle spreadsheets with custom applications that integrate BIM models, simulation software, and now, AI-generated code.

Together, they unpack one of today’s core tensions in engineering: how do we reconcile the reliability of deterministic systems with the flexibility of AI-driven workflows? If machines now help author our tools, where does responsibility live?

🔑 Key Takeaways:

• Why engineers struggle to trust AI and how deterministic code restores confidence
• How AI can generate tools, without becoming the tool itself
• What it really means to “automate the boring and engineer the awesome”
• Why Python is becoming the new engineering literacy
• How hybrid workflows blend human judgment with machine speed
• Why engineers must evolve from tool users into workflow architects

⏱ Timestamps

00:00 – When does automation become too unpredictable to trust?

02:15 – From aerospace engineer to code-curious

05:11 – What is VIKTOR and why build your own tools?

08:52 – Why engineers default to Excel under pressure

13:52 – “Automate the boring, engineer the awesome”

24:38 – Case study – automating the Fehmarnbelt Tunnel design

29:31 – Trusting AI vs trusting deterministic tools

35:59 – Unit tests and verification for engineering apps

41:16 – Connecting BIM, FEA and reports into one workflow

51:10 – When AI agents start acting like colleagues

56:18 – Where engineers should start with automation

59:43 – Outro

🔗 Guest:

Stijn Jansen CPO of VIKTOR

Website: ⁠https://www.viktor.ai/⁠

LinkedIn (Company): ⁠https://linkedin.com/company/viktor-ai⁠

LinkedIn (Stijn Jansen): ⁠https://www.linkedin.com/in/stijn-jansen-viktor⁠

VIKTOR Free account: ⁠https://www.viktor.ai/start-now⁠

🎧 Host & Links

LinkedIn: ⁠https://linkedin.com/in/owenrkeenan⁠

Spotify: ⁠https://open.spotify.com/show/7vcBFpl87NcqpLf0seIRaW

VoiceInspect.ai (AI for inspections): ⁠https://www.voiceinspect.ai⁠

👉 Subscribe to the Structural Intelligence Podcast on YouTube and Spotify to stay updated on new episodes.