Snigdh Sharma — Engineering Noteshttps://snigdhsharma.in/insightsPractical notes on data engineering, cloud platforms, and production AI.enTesting AI Agent Tool Authorization: Seven Reproducible Failure Caseshttps://snigdhsharma.in/insights/agent-tool-authorization-labhttps://snigdhsharma.in/insights/agent-tool-authorization-labWed, 14 Oct 2026 00:00:00 GMTA small open-source experiment comparing unchecked tool execution with tenant ownership, tool scope, exact approval, and identity checks. Includes an interactive lab and executable tests.Security EngineeringA Production AI Agent Checklist: Identity, Evaluation, and Recoveryhttps://snigdhsharma.in/insights/production-ai-agent-checklisthttps://snigdhsharma.in/insights/production-ai-agent-checklistWed, 14 Oct 2026 00:00:00 GMTA practical release checklist for agents that query enterprise data or call business APIs, with a worked example and explicit failure tests.AI EngineeringData Quality Contracts in Databricks: Beyond a Passing Pipelinehttps://snigdhsharma.in/insights/databricks-data-quality-contractshttps://snigdhsharma.in/insights/databricks-data-quality-contractsWed, 14 Oct 2026 00:00:00 GMTHow to define freshness, completeness, and failure handling before a dataset feeds a dashboard or AI assistant. Includes a small SQL validation example.Data EngineeringFrom Dashboards to Decisions: Why Analytics Maturity Is a Business Problem, Not a Tooling Onehttps://snigdhsharma.in/insights/from-dashboards-to-decisionshttps://snigdhsharma.in/insights/from-dashboards-to-decisionsMon, 01 Jun 2026 00:00:00 GMTMost organisations do not have a dashboard shortage — they have a decision shortage. A look at how to shift analytics investment from building reports to changing how teams actually decide.Data StrategyThe Lakehouse as a Business Operating Systemhttps://snigdhsharma.in/insights/lakehouse-as-business-operating-systemhttps://snigdhsharma.in/insights/lakehouse-as-business-operating-systemFri, 01 May 2026 00:00:00 GMTWhen the data platform becomes the single source of truth, it stops being IT infrastructure and starts being how the company runs. Lessons from building governed lakehouses on Azure.Cloud & AIPutting LLMs to Work Without Putting the Business at Riskhttps://snigdhsharma.in/insights/llms-in-the-enterprise-without-the-riskhttps://snigdhsharma.in/insights/llms-in-the-enterprise-without-the-riskWed, 01 Apr 2026 00:00:00 GMTAI agents promise huge productivity gains — but enterprise adoption lives or dies on governance, trust, and ROI. A pragmatic framework for separating the hype from the value.AI in the Enterprise