AI
Tango AI – AI Engine | Evolving Semiconductor Yield Analysis into Intelligent Decision-Making
Facing the dual pressures of increasing process complexity and rising test costs, Tango AI leverages its built-in AI modules to transform data into predictive and actionable solutions. Three core technologies deliver unprecedented efficiency gains and yield insights for our customers:
Wafer Map Classification
Traditional manual interpretation of wafer maps is prone to subjective errors.
Tango AI leverages deep learning models—such as CNN and ResNet—to automatically recognize defect patterns, enabling classification and anomaly alerting:
-
Automatically recognize common defect patterns (e.g., Edge Fail, Center Dot, Cluster)
-
Correlate with process stage data to enhance root cause identification accuracy
-
Establish anomaly trend monitoring to rapidly trigger process improvements
Result: Over 90% reduction in manual classification time, with anomaly detection accuracy exceeding 95%.
Test Time Reduction (TTR)
Intelligent test strategies accelerate the product validation process
By leveraging AI models to analyze historical test bin distributions, parameter sensitivity, and product stability, Tango AI helps customers optimize test time:
-
Automatically identify low-risk test items that can be removed or shortened
-
Recommend grouped testing strategies (Test Profiling) to reduce redundant validation
-
Simulate the impact of test reduction on yield and quality risk
Result: Test time can be reduced by an average of 20–40%, significantly shortening mass production cycles and lowering testing costs.
Yield Simulation
AI-Driven Yield Prediction and Virtual Validation Platform
Tango AI combines statistical modeling and machine learning to simulate the impact of various process parameters and test configurations on final shipment yield:
-
Conduct parameter sensitivity analysis based on WAT/CP/BUMP/FT data
-
Predict potential yield risks for new products or engineering lots
-
Recommend optimal process parameter combinations to achieve target yield (What-if Simulation)
Result: Predict yield trends before mass production to reduce product ramp-up risks.
Summary | From Data Analysis to AI-Driven Decision Making
Tango AI technology not only enhances yield analysis efficiency, but also provides:
✦Predictive Capability
Identify trends before anomalies occur
✦
Optimization Capability
Provide actionable recommendations on test strategies and process parameters
✦ Enhanced Risk Mitigation
Quantify anomalies and quality risks to enable reliable decision-making.
Tango AI Engine – Shaping the Future Core of Smart Manufacturing and Yield Management.
Tango AI × ChatGPT
Redefining the Language of Semiconductor Data Analytics
In the big data–driven semiconductor industry, data is not the problem—understanding it is the true competitive edge.
Tango AI, integrated with OpenAI’s ChatGPT and powered by Python, delivers a groundbreaking intelligent analytics platform that helps engineers rapidly identify yield anomalies and dramatically improves cross-department communication and decision-making efficiency.
Application
Wafer Data × Semantic AI: Instant Answers, Intelligent Understanding
Tango AI transforms massive datalogs from key semiconductor process stages—such as WAT, CP, FT, BUMP, and SLT—into semantically understandable, intelligent data assets.
Through structured API output and seamless Python integration with ChatGPT, complex data becomes instantly searchable, explainable, and visualizable.
Engineers can simply enter natural language queries like “Identify abnormal lots” or “Explain the high failure rate of a test item,” and the AI instantly responds with detailed analyses and visualizations—compressing what used to be time-consuming queries and interpretations into real-time insights.
Reduce Analysis Time by Over 70%: One-Click Insight Reports
Traditional yield analysis often requires hours of complex queries and manual comparisons.
Tango AI, integrated with ChatGPT, automatically generates reports based on real-time data trends—covering anomaly analysis, lot comparisons, and key parameter explanations.
Reports are available in PDF, HTML, and even email notification formats, providing rapid support for decision-makers.
Data Democratization: Seamless Cross-Department Communication, Data Everyone Can Understand
With ChatGPT natural language understanding, Tango AI serves as a bridge between departments.
No coding knowledge is required—just express your needs, and the system will deliver charts, explanations, and recommended actions.
This enables truly real-time, cross-functional collaborative analysis.
Accurate Simulation and Trend Forecasting: From Data Mining to Intelligent Recommendations
By combining Tango yield simulation engine with ChatGPT inference capabilities, users can simulate the impact of specific parameters on yield and analyze the potential gains from reduced test time.
With Tango AI, every data point becomes more than just history—it becomes a prediction for smarter decision-making.
Application Scenarios
✦ Wafer Yield Anomaly Monitoring
An engineer enters: "Please analyze products with FT yield below 95% this week."
ChatGPT automatically retrieves data from the Tango system, compares it with historical records, and outputs potential root causes (e.g., BUMP bonding issues, SLT fail bin shifts), along with suggested traceability paths across relevant process stages.
✦ Automated Generation of Test Analysis Reports
ChatGPT analyzes Datalog and generates visualizations such as histograms, trends, and CDF charts, explaining their significance in engineering terms.
The results are automatically formatted into a report and sent to the team.
✦ Natural Language Interactive Analysis
Users can input queries like: “What anomalies are present in this batch? How does it differ from last week?”
ChatGPT responds directly with data variances, anomaly distributions, and potential correlations.