Nidhi Sharma's contributions

Learn how to test AI database agents for cross-tenant data access risks using authorization checks, tenant isolation, and least-privilege security controls

Learn how to evaluate Azure Cosmos DB AI query assistance by measuring generated query accuracy, performance, request units, and cost efficiency.

Learn how to benchmark .NET Aspire 13.5 dashboard telemetry under large workloads and measure filtering, health checks, reconnections, and performance.

Learn how to test GitHub Copilot replacement models before September 2026 deprecations using repeatable benchmarks for quality, compatibility, latency, and cost.

Prepare for September 2026 GitHub Copilot model deprecations by identifying affected models, testing replacements, updating workflows, and avoiding development disruptions.

Learn how incorrect AI model routing affects cost, latency, response quality, and reliability using Microsoft.Extensions.AI routing capabilities and practical benchmarks.

Benchmark semantic routing and failover routing in Microsoft.Extensions.AI to compare model selection, reliability, latency, cost, and AI workload performance.

Compare AI coding agents with traditional chat assistants by measuring LLM workload complexity, tool calls, context usage, latency, token consumption, and execution patterns.

Benchmark GitHub Copilot agent tool-call retries and measure their impact on compute usage, latency, token consumption, reliability, and overall AI coding performance.

Compare semantic AI routing with static model selection in .NET by benchmarking latency, cost, routing accuracy, complexity, and overall application performance.

Build trace-based release gates for AI applications using evaluation metrics, execution traces, quality thresholds, regression detection, and automated CI/CD checks to prevent unreliable AI releases.

Learn how to measure AI agent cost attribution across enterprise projects by tracking model usage, token consumption, tool calls, workloads, teams, applications, and overall AI infrastructure spending.

Build secure hosted AI agents using isolated runtime environments with controlled compute, memory, filesystem access, permissions, network boundaries, and enterprise security controls.

Benchmark AI retrieval latency across Microsoft Fabric IQ and traditional vector stores by comparing response times, retrieval accuracy, indexing performance, scalability, and enterprise workload behavior.

Benchmark EF Core compiled queries on large production-style datasets to measure query execution time, allocations, throughput, and performance improvements under realistic workloads.

Build a tamper-resistant NuGet supply chain using package signing to verify package authenticity, protect build integrity, and reduce software supply chain risks.

Detect flaky .NET tests using historical failure analysis to identify recurring failures, distinguish regressions from instability, and improve automated testing reliability.

Build DNS failover detection with .NET health probes to identify resolution failures, service outages, endpoint changes, and application-level recovery conditions.

Build a DNS-aware failover client with .NET 11 that dynamically handles DNS changes, endpoint failures, recovery scenarios, and resilient service communication.

Benchmark C# async tail-await optimizations in .NET 11 to measure their impact on asynchronous execution, latency, throughput, allocations, and runtime performance.

Create a CI/CD security regression gate for .NET applications that validates runtime and framework updates before deployment and detects security issues early.

Benchmark .NET Lambda cold starts with Durable Workflows. Optimize performance, understand overhead, and make informed architectural decisions.

Secure AI agents by testing unauthorized data access. Focus on authorization outside the model, tool misuse, and indirect injection.