Uploaded August 2026 | Updated September 2026, 3 weeks ago
AI applications depend on scalable and reliable data platforms.
This session explores how Microsoft Fabric enables modern data engineering workflows through lakehouse architecture, data ingestion, orchestration, Spark-based processing, analytics, and governance.
Attendees will learn how to build strong data foundations to support analytics, machine learning, and generative AI workloads.
Key takeaways:
- Modern lakehouse architecture
- Building scalable data pipelines
- Spark and Fabric integration
- Data engineering best practices for AI workloads
00:00 Welcome & Housekeeping
01:41 Speaker Introduction & Background
03:23 AI Language Bias and the English-First Problem
06:19 Why Multilingual AI Matters
08:59 Case Study: Document Extraction Across Languages
11:58 Challenges in Non-English NLP
14:37 Architecting Multilingual AI Solutions
19:07 Common Pitfalls: Language Identification & Data Conversion
25:07 Processing Non-Latin Characters
27:40 Evaluating Multilingual AI Systems
29:54 Key Takeaways & Transition to Q&A
32:15 Introduction to FLAIR and DRI Copilot
36:04 Why Retrieval Systems Fail Over Time
38:20 Historical Analysis of User Queries
40:50 FLAIR: Using Feedback to Improve Retrieval
43:21 Offline Preprocessing & Signal Generation
46:28 Online Retrieval and Reranking
50:31 Computing Vote Scores from User Feedback
56:04 Evaluation Methodology & Results
1:01:45 Audience Q&A
1:03:50 Closing Remarks & Survey
📌 This is part of a series, learn more here: https://aka.ms/ProdReadySystems/series
[eventID:27360]
AI applications depend on scalable and reliable data platforms.
This session explores how Microsoft Fabric enables modern data engineering workflows through lakehouse architecture, data ingestion, orchestration, Spark-based processing, analytics, and governance.
Attendees will learn how to build strong data foundations to support analytics, machine learning, and generative AI workloads.
Key takeaways:
- Modern lakehouse architecture
- Building scalable data pipelines
- Spark and Fabric integration
- Data engineering best practices for AI workloads
00:00 Welcome & Housekeeping
01:41 Speaker Introduction & Background
03:23 AI Language Bias and the English-First Problem
06:19 Why Multilingual AI Matters
08:59 Case Study: Document Extraction Across Languages
11:58 Challenges in Non-English NLP
14:37 Architecting Multilingual AI Solutions
19:07 Common Pitfalls: Language Identification & Data Conversion
25:07 Processing Non-Latin Characters
27:40 Evaluating Multilingual AI Systems
29:54 Key Takeaways & Transition to Q&A
32:15 Introduction to FLAIR and DRI Copilot
36:04 Why Retrieval Systems Fail Over Time
38:20 Historical Analysis of User Queries
40:50 FLAIR: Using Feedback to Improve Retrieval
43:21 Offline Preprocessing & Signal Generation
46:28 Online Retrieval and Reranking
50:31 Computing Vote Scores from User Feedback
56:04 Evaluation Methodology & Results
1:01:45 Audience Q&A
1:03:50 Closing Remarks & Survey
📌 This is part of a series, learn more here: https://aka.ms/ProdReadySystems/series
[eventID:27360]
