从 Text2SQL 到权益核销：业务数据智能查询 Agent 工程复盘
从 DDL、指标口径和 SQL 评估，到 MQ 幂等、缓存一致性和补偿机制，拆解可靠数据 Agent 的完整链路。
Read engineering reviewData Products
A topic hub for analytics systems, AI data products, data engineering, experimentation, decision support, monitoring, and product workflows.
Beyond dashboards
Beyond dashboards
A data product is a repeated workflow that turns data into a decision, action, or operating habit. It needs a named user, a decision cadence, trusted definitions, and a feedback loop.
AI data products
AI data products
AI data products need traceable model outputs, visible uncertainty, review paths, and enough metadata to understand which data, prompt, schema, and reviewer decision produced an output.
What to monitor
What to monitor
Track freshness, schema changes, missing values, usage, adoption, overrides, drift, correction rate, experiment impact, and recurring interpretation questions.
Topic hub
Data product writing
Data products are repeated decision workflows, not just dashboards. They need a user, a cadence, trusted definitions, instrumentation, and a feedback loop that shows whether the product changes behavior or improves decisions.
For AI data products, the bar is higher: outputs need provenance, uncertainty, correction paths, and metrics that reveal adoption and quality. The writing here connects experimentation, causal thinking, analytics, platform work, and product framing so data systems become useful operational tools.
This section follows the starter-blog pattern of metadata-driven topic pages: each card shows a canonical article link, summary, date, reading time, and tags. The page is a hub, not a duplicate post archive. For search-driven practical guides, start with A/B testing sample size in Python, use the A/B test sample size calculator, then continue to causal inference for product analytics and model calibration in Python.
Building a Practical Enterprise Data Agent with Microsoft Fabric Data Agent and Azure AI Foundry
A practical architecture for governed data agents that connect Fabric data assets, ontology, MCP endpoints, Foundry orchestration, and application workflows.
Read architecture guideA 90% Accurate Model That Still Loses Money
Why high-accuracy churn prediction can improve renewal rate while losing revenue, and how uplift thinking targets incremental renewals instead of risk scores.
Read articleBeyond A/B Testing: Uplift Modeling in Industry
A practical guide to CATE, meta-learners, Qini, AUUC, decile lift, online experiments, and incremental ROI.
Read articleCausal Inference for Product Analytics
How product teams connect experiments, observational data, ATE, CATE, uplift, guardrails, and decision quality.
Read guideDatabricks Lakehouse Guide
Delta Lake, MLflow, Unity Catalog, and data platform implementation notes.
Read articleA/B Test Engineering Guide
Feature gates, experiments, reliable launch measurement, and experimentation infrastructure.
Read articleA/B Testing Sample Size in Python
Python sample size calculations, power, minimum detectable effect, guardrails, and experiment readiness checks.
Read sample size guideModel Calibration in Python
Reliability diagrams, expected calibration error, confidence bins, thresholds, and monitoring checks for trustworthy ML.
Read calibration guideKey Statistical Tests for Survey Analysis
A practical guide to choosing statistical tests and interpreting experiment-style evidence.
Read articleHealthcare AI Analytics
Applied AI and analytics framing for healthcare operations and decision support.
Read articleFor future notes on applied AI systems, evaluation, data products, and product workflows, subscribe to Yangming Li's Newsletter.
Subscribe for updates