Course Objective & Scope
Across a platform with hundreds of tables, writing manual assertions for every column is impossible—and upstream delays or sudden volume drops often slip right past static unit tests. This training adapts software observability principles to data and AI pipelines so teams can catch unknown-unknown data issues proactively.
Lightweight metadata-driven monitoring, deep column profiling, time-series anomaly detection for freshness and volume, distribution drift metrics, open-source tooling (such as Elementary) and incident management workflows are covered hands-on. Participants design a monitoring setup, data SLI/SLO targets and an incident runbook for their own critical datasets.
Who Should Attend?
- Data platform, data engineering and DataOps teams
- MLOps and LLMOps engineers monitoring live data feeds for ML and RAG systems
- Analytics leads responsible for data reliability and SLA management
- Prerequisite: SQL, warehouse/lakehouse concepts and basic orchestration knowledge
Key Takeaways
- Monitor table freshness and volume at near-zero query cost using information schemas and table metadata
- Configure dynamic anomaly thresholds that account for seasonality and weekday/weekend patterns
- Detect sudden NULL spikes, cardinality shifts and numerical distribution drift
- Define data SLIs/SLOs for tier-1 tables and route high-signal alerts without causing alert fatigue
- Run impact analysis, root-cause triage and blameless post-mortems using a structured data incident runbook
Curriculum
Data Testing vs Data Observability
- Difference between verifying known rules (testing) and discovering unknown anomalies (observability)
- Data reliability metrics: reducing MTTD (time to detect) and MTTR (time to resolve)
Low-Cost Metadata Monitoring: Freshness, Volume and Schema
- Using table logs, Iceberg/Delta snapshot metadata and information_schema
- Catching latency and row-count anomalies with seasonal time-series baselines
Deep Profiling and Distribution Drift
- Tracking NULL rates, uniqueness, quantiles and categorical frequency shifts
- Controlling compute cost with sampling and partition-aware scans
Data SLIs/SLOs and High-Signal Alerting
- Setting freshness and completeness SLOs for tier-1 analytics and AI tables
- Grouping correlated downstream failures and routing alerts to Slack/PagerDuty
Data Incident Response and Root-Cause Triage
- Separating code regressions, upstream data changes and infrastructure failures
- Safe backfill strategies, stakeholder communication and blameless post-mortems
Other Courses in This Category
-
High-accuracy extraction of structured text and metadata from complex PDF tables, scanned forms, audio recordings and images using OCR, Docling, Whisper and vision-language models.
- Live Course
- 5 modules
-
Modular SQL transformations with dbt Core/Cloud, staging-intermediate-mart layers, Jinja macros, incremental models, snapshots, testing and CI/CD workflows for data and analytics engineers.
- Live Course
- 5 modules
-
Authoring production DAGs with Apache Airflow: TaskFlow API, dynamic task mapping, dataset/asset-aware scheduling, isolated execution on Kubernetes, error handling and AI/RAG pipeline orchestration.
- Live Course
- 5 modules
-
Dirty Data, Clean CV
Advanced workshop on applying regression to messy real-world data, presenting model metrics to non-technical stakeholders, acing ML technical interviews, and building an impactful Data Science CV and GitHub portfolio.
- Live Course
- 4 modules
-
Warehouse reliability on object storage: Apache Iceberg table internals, metadata tree, hidden partitioning, time travel, REST Catalog, multi-engine querying (Trino, Spark, DuckDB) and table maintenance.
- Live Course
- 5 modules
-
Centralising enterprise metrics and dimensions with dbt Semantic Layer (MetricFlow) and Cube so BI tools, Text-to-SQL systems and analytical AI agents return consistent numbers.
- Live Course
- 5 modules
Get in touch
