Jedat Inc. · TSE:3841
Jedat — Pollen Metrology — Pollen Webinar #3 - Collaborative AI for Defectivity (February 2026)
34 slides from Jedat's conference presentation, Feb 16, 2026
Slide 1: Webinar #3 Collaborative AI Software Layer for Defectivity
Webinar #3 Collaborative AI Software Layer for Defectivity 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 1
Slide 2: Speakers
Speakers 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 2
Slide 3: Webinar #3 – Agenda
Webinar #3 – Agenda Duration: 45min Collaborative AI Software Layer for Defectivity 30min Introduction to Collaborative AI used in industrial environments Overview of how state of the art AI and Data Science techniques can automate defectivity on Microscopy images boosting defect detection and inspe
Slide 4: 1. Context of Data Analytics for Semiconductor Defect Inspection
1. Context of Data Analytics for Semiconductor Defect Inspection 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 4
Slide 5: Innovative Materials Ecosystem Data Flow
Innovative Materials Ecosystem Data Flow Traditional Engineer-Driven Materials Development From fragmented data to actionable process intelligence Without shared analytics, engineers cannot fully leverage either wheel Simulation R&D and Production line Process Process Process Process parameters mode
Slide 6: Semiconductor Fabs Global
Semiconductor Fabs Global Three departments. One process reality. Inspection Everywhere Disconnected analytics prevent cross-department learning on inspection IC Makers GLOBAL QUALITY & RELIABILITY Customers & Fabless • Failure • Root cause analysis Updates models / Recipes / analysis • Reliability
Slide 7: Artificial Intelligence has reached industrial maturity for process control
Artificial Intelligence has reached industrial maturity for process control Accelerate inspection capabilities to maximize yield with a scalable solution Hardware + Software + Data Intelligence 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 7
Slide 8: Augmenting Engineers With a Continuous AI-Driven Development Loop
Augmenting Engineers With a Continuous AI-Driven Development Loop But maturity does not come from AI alone — it comes from how AI is integrated into engineering loops Simulation R&D and Production line Process Process Experts Process Process parameters modelling DOE tool logfile knowledge Wheel of f
Slide 9: Semiconductor Manufacturing Fab Software Stack
Semiconductor Manufacturing Fab Software Stack Example of data flow management • Black Box • Poor Flexibility Data lives in five layers — engineers must bring it together • No external contribution YMS Yield Analytics Yield Management System ADC Data quality & Defect Intelligence Automatic Defect Cl
Slide 10: Collaborative AI
The sixth software layer - an AI orchestration layer above MES, YMS, ADC, and control systems • Open Platform • Full Flexibility • External contribution YMS SOFTWARE PLATFORM Yield Management System COLLABORATIVE AI ADC Automatic Defect Classification MES Manufacturing Exec. System FDC SPC R2R Contr
Slide 11: 2. Our vision and mission
2. Our vision and mission 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 11
Slide 12: Pollen - Collaborative AI for Defect Inspection
Pollen - Collaborative AI for Defect Inspection AI-Driven Software for Next-Generation Manufacturing Bridging Data Silos in Material Science with Collaborative AI for Yield Enhancement • Who we are Founded in 2014, Pollen is a software company based in Grenoble, France, with a team of 20 experts ded
Slide 13: The Sixth Layer: Orchestrating All AI Technologies Seamlessly
The Sixth Layer: Orchestrating All AI Technologies Seamlessly Smart3 – AI Orchestration Layer connecting R&D, Manufacturing & Quality GLOBAL QUALITY & RELIABILITY IC Makers & Fabless • Failure • Root cause analysis Customers analysis • Reliability monitoring Updates models / Recipes / • Failure proc
Slide 14: Smart3 seamlessly integrates with tool software
Smart3 seamlessly integrates with tool software Smart3 – Secured Application Software Layer for Tool Suppliers TOOL SUPPLIER ECOSYSTEM TOOL APPLICATION SOFTWARE SMART3 AI Capabilities - SECURED APPLICATION SOFTWARE LAYER - - inside Smart3 - • Metrology / Inspection logic (Embedded alongside tool sof
Slide 15: Example of Smart3 Integration for Fabs or Tools
Example of Smart3 Integration for Fabs or Tools Use the the SDK & the API or 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 15
Slide 16: Comprehensive Smart3 Software Suite to Boost Semiconductor Production
Comprehensive Smart3 Software Suite to Boost the Production of Semiconductor Devices Metrology Inspection Two adoption models Focus of this Webinar 1. Turnkey AI-Driven Process Control Software Process Optimization 2. Smart3 as a Low-Code SDK Framework to accelerate 10×+ internal application develop
Slide 17: 3. Industry challenges and AI for Defectivity
3. Industry challenges and AI for Defectivity 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 17
Slide 18: Problems to solve
Problems to solve Focused on Inspection Time-pressure - Long inspection time - Costly expert time (engineers, data scientists) Process complexity - Defects can have few to no examples - Image artifacts must be separated from defects Low ROI from data - Data underused - Manual inspection reducing dat
Slide 19: Defectivity: Where Yield Is Won or Lost
Defectivity: Where Yield Is Won or Lost Focused on Inspection Time-pressure Supervised Classification - Long inspection time - Costly expert time (engineers, data scientists) Process complexity - Defects can have few to no examples Image Segmentation - Image artifacts must be separated from defects
Slide 20: Supervised image classification
Supervised image classification Automated Image Classification: • Provide class probabilities based on the full image content • Allows to process images with different algorithm/measurement depending on the class • Ultra fast annotation Gains: + Time gain + Can discriminate subtle differences + Auto
Slide 21: Image segmentation
Image segmentation Image segmentation to automate metrology: • Object detection to find rough location • Instance segmentation to delineate precisely objects • Semantic segmentation to delineate materials Challenge: • Integration of these tools with no code for the users and minimal amount of annota
Slide 22: Anomaly detection & Unsupervised Classification
Anomaly detection & Unsupervised Classification Anomaly detection: • Define positive objects • Detect defects that are different from positive objects • Obtain a rough localization of anomalies Challenge: • Detect unseen defects without flooding with False Positive Gains: + Fast annotation + Detect
Slide 23: Clustering
Clustering Clustering: • Separate various objects in classes automatically • Explore larger quantity of data faster Challenge: • Efficient classification without user inputs Gains: + Automation of analysis + Preprocessing of large amount of data + Synergy with Anomaly Detection 16/2/2026 POLLEN WEBI
Slide 24: 4. Use Cases & Demo
4. Use Cases & Demo 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 24
Slide 25: Demo plan
Demo plan 1. Collaborative Platform ➔ Example Integration Python sklearn library 2. Annotations & Deep learning for Inspection 3. Unsupervised: Anomaly Detection & Clustering 4. Collaborative Workflow Strategy & HVM Deployment 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 25
Slide 26: 1- Collaborative Platform Example Video
1- Collaborative Platform Example Video Plugin Manager Ease of use internal library to customize Smart3 platform without any software developement https://youtu.be/UJ0uPbH8o_c 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 26
Slide 27: 2- Annotation & Deep Learning for Inspection
2- Annotation & Deep Learning for Inspection https://youtu.be/ts8Q1vinyS4 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 27
Slide 28: 3- Implementation for HVM – Example Video
3- Implemention for HVM – Example Video Use the API https://youtu.be/a2GJeazDVZo 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 28
Slide 29: 4- Example of integration at the Manufacturing level
4- Example of integration at the Manufacturing level to increase Inspection Hardware ROI and Robustness Process # Yield THERMAL EPITAXY TREATMENT YIELD ENHANCEMENT Slip Lines Analysis can be: PWG* Computing • CNN, R-CNN • Third Party softwares • Internal Software Code Post computing User results dat
Slide 30: 5- Collaborative Workflow Strategy Videos
5- Collaborative Workflow Strategy Videos SmartOEM3: Collaborative AI at the Service of Supplier IP Wavefront Phase Imaging https://wooptix.com/ https://youtu.be/f9SHO3FzMoo?si=s9fCFa1Q6Ia9BsKO Atomic Force Microscope https://www.infinitesima.com/ https://youtu.be/TzAr2V1mmkg 16/2/2026 POLLEN WEBINA
Slide 31: 5. Next Webinars, Q&A
5. Next Webinars, Q&A 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 31
Slide 32: Next Webinars
Next Webinars Webinar 1 – Dec 2 (2025): Collaborative AI for Next-Gen Semi Manufacturing - A Software Platform Revolution Webinar 2 – Feb 3 (2026): Collaborative AI Software Platform for Metrology Webinar 3: Monday Feb. 16: Collaborative AI for Automatic Defect Classification 2026 – Q1 Webinar 4: Tu
Slide 33: Any question?
Any question? Seminar slides access , please send an email to [email protected] 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 33
Slide 34: Thank you
Thank you 16/2/2026 POLLEN WEBINAR 3 - COLLABORATIVE AI FOR DEFECTIVITY 34
Pollen Metrology — Pollen Webinar #3 - Collaborative AI for Defectivity (February 2026) on 6ix