VGhellovg.sys
HomeProjectsBlogResume
HELLOVG SYSTEMS // CV TERMINALBUILD 86.09 // ONLINE
>_hellovg.sys/Vimal Govind Markkasseri

Enterprise software engineer specializing in reliable ERP systems, .NET modernization, SQL Server, and practical AI-assisted workflows.

All systems operational · Manama, Bahrain GMT+3

Navigation

  • >Home
  • >Projects
  • >Blog
  • >Resume
  • >RSS Feed

Connect

GitHubLinkedInXInstagram[email protected]

© 2026 Vimal Govind Markkasseri. Built with Next.js & Cloudflare Pages.

Press CtrlJ anytime for quick commands

~/blog

Notes on software engineering, architecture, and practical AI.

Field notes from building and modernizing enterprise systems with .NET, SQL Server, web platforms, and AI-assisted workflows.

Showing 9 of 38 articles

Oct 30, 2026
6 min read

Beyond Autocomplete: Designing Multi-Agent Coding and Verification Workflows in Enterprise Teams

AI coding has evolved from inline tab-completion to autonomous multi-agent engineering workflows. How repository context standards like AGENTS.md, specialized subagents, and deterministic test gates protect enterprise codebases while accelerating delivery.

Developer ToolsAI AgentsEngineering LeadershipTestingEnterprise
Oct 28, 2026
6 min read

.NET 10 LTS and EF Core 10 in Enterprise ERP: Steady-State Gains, Outer Joins, and Migration Discipline

.NET 10 delivers long-term support with tangible runtime efficiency for 24/7 ERP workloads. A look at what actually moves the needle in production: native LINQ LeftJoin in EF Core 10, JSON mapping, GC steady-state memory behavior, and realistic LTS migration criteria.

.NETC#EF CoreEnterpriseArchitecturePerformance
Oct 22, 2026
6 min read

Small Reasoning Models and Edge Inference: Slashing Enterprise Token Budgets and Latency

Routing every internal request to massive frontier models is an expensive antipattern. By deploying Small Reasoning Models (SLMs) and edge inference via Cloudflare Workers AI and local models, enterprise systems can handle 80% of classification, triage, and grounding checks with sub-100ms latency and minimal cost.

Edge AICloudflare Workers AISLMAI ArchitectureCost Optimization
Oct 18, 2026
6 min read

SQL Server 2025 Native Vector and JSON: Pragmatic ERP Search Without a Dedicated Vector DB

SQL Server 2025 introduces native binary JSON storage and vector indexing. Before adding another database to your infrastructure, evaluate whether keeping transactional data, tenant security, and semantic search unified in SQL Server solves your ERP retrieval needs.

SQL ServerDatabaseAIArchitectureERPPerformance
Oct 12, 2026
6 min read

Computer-Use & Browser Agents in Enterprise ERP: Where UI Automation Belongs and Where It Breaks

Visual computer-use and browser agents promise to automate legacy vendor portals and desktop clients that lack modern APIs. But relying on screenshot loops for critical business workflows introduces non-determinism, cost, and security risks. Here is an enterprise triage framework.

Computer UseAI AgentsEnterpriseRPAMCPSecurity
Oct 8, 2026
6 min read

MCP Tools in Enterprise .NET: Enforcing Boundaries, Tenant Isolation, and Human Confirmation

Connecting an LLM agent to ERP APIs through the Model Context Protocol is deceptively easy. Keeping the agent from posting unauthorized general ledger entries or leaking cross-tenant data requires strict tool boundaries, claims propagation, and human-in-the-loop workflows.

MCP.NETAI AgentsEnterpriseSecurityArchitecture
Oct 4, 2026
5 min read

The Microsoft Agent Framework: What the Semantic Kernel and AutoGen Convergence Means for .NET Teams

Microsoft has unified the enterprise stability of Semantic Kernel with the multi-agent orchestration of AutoGen into the Microsoft Agent Framework. Here is an architectural look at how the converged runtime sits on Microsoft.Extensions.AI and how to design stateful, multi-agent ERP workflows.

Microsoft Agent FrameworkSemantic KernelAutoGen.NETAI AgentsEnterprise
Sep 24, 2026
3 min read

Evaluating a Grounded AI Assistant Before Changing Its Model

A model upgrade needs more than a few convincing answers. Build a small evaluation set that separates retrieval failures, unsupported claims, scope handling, latency, and cost.

AIEnterpriseProductionDeveloper Tools
Aug 21, 2026
3 min read

Retries and Idempotency: Protecting Business Operations

A timeout does not tell you whether a write succeeded. Design stable operation identities, bounded retries, and recovery paths before automatically repeating business requests.

APIArchitectureERPProduction