Metal additive manufacturing is a process of exquisite, expensive uncertainty. A single aerospace bracket can take days to print, only to be scrapped after an X-ray reveals a hairline crack or a subsurface pore. The cost of qualification, the exhaustive process of proving a printed part meets specification, often outweighs the cost of the metal and the machine time. NexusAM, a London-based startup, is building software that aims to collapse that uncertainty by predicting flaws while the laser is still melting powder.
The company's core product is an AI-powered process monitoring and qualification platform. It ingests data from in-situ sensors on metal 3D printers,thermal cameras, photodiodes, and others,and applies proprietary machine learning models to detect anomalies indicative of defects. The goal is to provide a real-time quality score and a predictive flag, allowing an operator to pause a failing build before wasting more material and time, and to generate a traceable digital thread for part certification.
The qualification bottleneck
For industrial users in aerospace, medical, and energy, the promise of metal 3D printing is design freedom and lightweight, complex parts. The reality is a validation gauntlet. Each new part design, and often each new build location on a printer's plate, requires a fresh qualification campaign. This involves printing, destructively testing, and meticulously documenting hundreds of samples to build a statistical process control model. It's a barrier that limits metal AM to high-value, low-volume production. NexusAM's bet is that AI can learn the signatures of a good print from sensor data, correlate those signatures with final part quality, and ultimately certify parts based on their digital manufacturing record rather than solely on post-build inspection [NexusAM, 2024].
The team and the technical wedge
Founders Sebastian Larsen and Shaaz Ghouse are building on research from Imperial College London, where the company was incubated and later selected for the Imperial-led AI SuperConnector Accelerator [Entrepreneurship Centre blog, 2026]. Larsen, the CTO, brings a background in applied machine learning from roles at Onfido and Faculty, with a focus on deep learning and probabilistic modeling [LinkedIn, 2024]. Ghouse, the CEO, has experience across additive manufacturing and medical devices from previous roles at OSSTEC and Stanmore Implants [RocketReach, 2026]. This pairing suggests a deliberate balance between the AI stack and the domain-specific understanding of metal AM workflows and regulatory requirements.
The technical differentiation, according to the company, lies in its proprietary AI stack. Rather than just monitoring basic thermal signatures, the platform is designed to model the complex, multi-physics interactions of the laser-powder melt pool. The aim is to predict not just gross failures, but subtle material inconsistencies,like porosity or lack-of-fusion defects,that can measure in the tens of microns but compromise structural integrity. Creator Fund, the startup's early backer, highlighted the company's focus on "the prohibitive cost of part qualification" as its primary wedge [The Creator Fund, 2024].
A crowded field of sensors
NexusAM is not alone in trying to instrument the metal AM process. The competitive landscape includes several established and emerging players, each with a slightly different technical approach or market focus.
| Company | Primary Focus | Approach |
|---|---|---|
| Addiguru | In-situ monitoring & analytics | Software platform for process monitoring and data management. |
| Phase3D | In-situ distortion monitoring | Fringe projection system to measure part deformation during printing. |
| Additive Assurance | Quality assurance for aerospace | Acoustic monitoring technology to detect anomalies. |
| Senvol | AM database & analytics | Materials and process database software for qualification. |
| DONAA | Process monitoring | Multi-sensor monitoring systems for laser powder bed fusion. |
NexusAM's stated edge is the depth of its AI prediction layer, moving from monitoring to prescriptive quality assurance. However, the space is capital-intensive, requiring deep R&D and lengthy sales cycles to penetrate risk-averse industrial customers. The lack of publicly disclosed pilot customers or funding amount beyond Creator Fund's involvement indicates NexusAM is still in the early validation phase.
The path to industrial floors
The immediate technical challenge is moving from lab data to production noise. Printer models vary, sensor calibrations drift, and material batches differ. An AI model trained on one machine with one alloy may not generalize. The company's near-term roadmap likely involves forging deep partnerships with a handful of lead customers to refine its models on their specific production lines.
Financially, the model is classic enterprise SaaS, but adoption hinges on proving a clear return on investment. The value proposition must be quantified in hard terms: reduction in scrap rate, acceleration of qualification timelines, and the avoided cost of failed builds. For a sector where a single scrapped part can represent tens of thousands of dollars in wasted superalloy powder, the savings potential is significant if the prediction accuracy is high enough.
Technical breakdown. The system's effectiveness will be measured by two key metrics: false positive rate and detection latency. A high false positive rate, where the system halts good prints, destroys the economic benefit. Detection latency is equally critical; spotting a flaw at 80% build completion saves some cost, but spotting it at 20% saves far more. The real test is whether the AI can identify the precursor signatures of a defect that only manifests layers later.
The sober assessment is that success depends on data density and domain specificity. The AI is only as good as the training corpus of flawed prints, which are expensive to produce intentionally. Furthermore, scaling requires the software to become printer-agnostic, integrating with a wide array of sensor suites and control systems without becoming a custom integration project for every client. If NexusAM can navigate these hurdles and demonstrate reliable performance in a production environment, it could become a critical piece of infrastructure for the next generation of digital manufacturing.
Sources
- [NexusAM, 2024] NexusAM Website | https://www.nexus-am.com
- [The Creator Fund, 2024] NexusAM Case Study | https://thecreatorfund.com/case-studies/nexusam/
- [LinkedIn, 2024] Sebastian Larsen Profile | https://www.linkedin.com/in/sebastian-larsen-60102748
- [RocketReach, 2026] Shaaz Ghouse Profile | https://rocketreach.co/shaaz-ghouse-email_239080677
- [Entrepreneurship Centre blog, 2026] Harnessing AI and 3D Printing for Digital Manufacturing | https://entrepreneurship.blog.jbs.cam.ac.uk/2024/09/17/harnessing-ai-and-3d-printing-for-digital-manufacturing/
- [CB Insights, 2026] Top Addiguru Alternatives, Competitors | https://www.cbinsights.com/company/addiguru/alternatives-competitors