Let's Combinate · 15 Oct 2025 · 22m
Ben Locwin, a Healthcare Executive and Quality SME, joins host Subhi Saadeh to discuss integrating AI into Good Manufacturing Practices (GMP) within pharmaceutical and medical device manufacturing. They explore current applications like AI-assisted CAPA and deviation management, while highlighting the regulatory challenges of algorithmic opacity in Process Analytical Technology (PAT) and continuous manufacturing. The conversation emphasizes the critical role of Design of Experiments (DOE) in training AI models and using yield calculations as an early warning system for process discrepancies in combination products.
Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.
Quality & test
AI analyzes historical deviation and complaint data to suggest likely root causes and corrective actions for investigation reports.
“I've seen some use cases of deviation management, where the AI is analyzing historical deviations, could be on the device side, complaints, and kappas, and suggesting likely root causes and corrective actions.”
Speaker 2 · Hear it at 0:46
Manufacturing
AI models analyze real-time process and test data from Process Analytical Technology (PAT) to recommend or automate batch release decisions in continuous manufacturing.
“having AI models analyze real-time process and test data, and then recommend or automate batch release decisions. And in particular, those in continuous manufacturing environments to say … this batch is okay to be released…”
Speaker 2 · Hear it at 4:41
Manufacturing
AI helps interpret Process Analytical Technology (PAT) data, such as NIR spectroscopy, to control quality attributes in real time by layering multiple process parameters.
“another AI use case is helping to interpret PAT data. So like if you have NIR spectroscopy data, and you're trying to control your quality attributes in real time based on that … the model could support layering five, 10, 15, 20 other attributes in there…”
Speaker 2 · Hear it at 9:15
large enterprises
Other shows on the same ground, closest first.
Host John Spear of Greenlight Guru interviews Devon Campbell, founder of Prodct LLC, and Anthony Habayeb, co-founder and CEO of Monitaur.ai, about the challenges and frameworks for integrating artificial intelligence (AI) and machine learning (ML) into medical devices. The discussion centers on 'Machine Learning Assurance,' a process for auditing and monitoring ML models to ensure safety and effectiveness in regulated environments, contrasting it with traditional Quality Management Systems (QMS). The guests share insights from FDA workshops and guidance, emphasizing the need for a 'culture of quality control' and Algorithmic Change Protocols (ACP) to manage the risks of evolving algorithms in Software as a Medical Device (SaMD).
Mike Drues, an industry veteran with over 30 years of experience, joins host Etienne Nichols to discuss the dangers of 'boneheaded' regulatory mistakes in MedTech, specifically focusing on the non-delegable responsibility of manufacturers. They analyze the first FDA warning letter issued for GMP violations related to AI, where a company used AI to generate specifications without process validation, arguing that AI cannot replace human judgment or the Quality Unit. The conversation extends to the debate on individual accountability in warning letters, the distinction between wellness and regulated devices, and new FDA programs like the TEMPO pilot and Breakthrough Device Designation (BDD).
Etsy Nichols discusses the evolving FDA regulatory landscape for AI-enabled medical devices, highlighting the establishment of the Digital Health Center of Excellence and the shift toward rigorous documentation over algorithmic sophistication. She outlines five guiding principles for AI/ML-based Software as a Medical Device (SaMD) and emphasizes the critical importance of Predetermined Change Control Plans (PCCPs) in bridging US and EU regulatory requirements. Nichols argues that QA/RA professionals must combine regulatory fluency with AI literacy to navigate these complex systems effectively.
Tom Rynkiewicz, VP of RA/QA at Raland Compliance Partners, joins host John Spear to discuss the top compliance issues plaguing medical device manufacturers, specifically CAPA, complaints, and design controls. Rynkiewicz argues that these recurring FDA observations stem from flawed internal audit practices that audit isolated quality system components rather than the entire system holistically. He highlights how human behavior and corporate KPIs, such as tracking CAPA turnaround times instead of effectiveness, prevent companies from achieving a true quality culture with zero complaints. The conversation emphasizes the need for system-level auditing and transparency to prepare for potential FDA shifts toward stricter Good Manufacturing Practices (GMP) inspections.
Host Etienne Nichols analyzes the FDA's warning letter to Purolea Cosmetics Lab, arguing that the citation was not a crackdown on AI but a failure of quality system oversight and process validation. He explains that while Purolea used AI agents to draft specifications, the core violation was the lack of human review and accountability under 21 CFR 211.22. Nichols outlines a three-part validation framework for AI/ML-based Software as a Medical Device (SaMD), emphasizing the separation of training, validation, and holdout test data. He concludes by listing essential regulatory standards and five concrete questions to assess if a quality system is ready for FDA AI inspections.
The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Let's Combinate. The episode itself belongs to its publisher.