Senior Software Engineer & Solutions Architect .NET modernization + AI / agentic systems · Over 20 years Enterprise .NET & cloud architecture, and an independent founder shipping AI-native products end to end.
I'm a senior software engineer and solutions architect with over 20 years building enterprise-grade systems across .NET, cloud, and integration-heavy environments — and the founder of Carborundum AI, where I design and ship AI-native products from zero.
My work spans two complementary tracks. The first is application modernization: legacy .NET decomposition, migration to .NET Core / .NET 10, Clean Architecture and modular-monolith refactoring, Azure-native APIs, and enterprise integrations (Salesforce, document-management platforms, AI-assisted processing). The second is applied AI engineering: multi-agent orchestration on the Claude API, RAG and memory pipelines, and agentic workflow design — built from first principles, not wrapper libraries.
I own complex architecture end-to-end and move from architecture to working code to deployed system without a committee. Most recently I delivered cloud-native .NET 10 microservices and AI-assisted insurance workflows for Lockton on a Perficient app-modernization engagement; independently I've built and shipped two complete products — Carborundum AI and MiChoice — solo.
What I bring to a team:
- Application modernization from assessment to delivery — legacy decomposition, microservices and modular-monolith refactoring, Azure-native APIs
- Applied AI engineering — multi-agent orchestration on the Claude/Anthropic API, RAG/memory pipelines, agentic workflow design
- Full-stack product delivery — Next.js/React front ends, .NET/Blazor services, Stripe, CI/CD, deployed and running in production
- Enterprise integration depth — Salesforce (Forza.Net/SOQL), M-Files, document processing, distributed caching
- The ability to go from architecture to deployed system independently
| Project | What it does | Tech stack | Link |
|---|---|---|---|
| Carborundum AI | AI-native field-operations platform for trades businesses (HVAC, plumbing, electrical), built from zero. A Next.js platform (25+ routes, subdomain routing, Vercel auto-deploy) over a modular ERP suite (Manifold) covering job costing, scheduling, CRM, and invoicing; a multi-agent AI layer integrating the Claude API, ElevenLabs voice, and HeyGen avatars under a written operating constitution; Stripe payment flows; and a React Native companion field app (Titanium). WCAG 2.1 AA across the product surface. | Next.js, React, Node, Claude API, React Native, Stripe, Vercel | carborundum.ai |
| MiChoice | Multi-module K-12 school-nutrition SaaS suite, built end-to-end. Multi-tenant modules for district back-office (miComs), parent free-&-reduced applications (miFAR), cashier POS (miMeal), and time-clock (miTimes), plus an AI advisory insight layer (miAgentic) that switches to the Anthropic API for recommendations. Deployed on Railway. | C#/.NET, Blazor Server, EF Core, Next.js/React, Anthropic API, Railway | private — demo on request |
| A.S.K (Agent Skills Kernel) | A versioned, composable library of AI agent capabilities. Think .dll files for AI systems — write a capability once (GitHub push, Telegram notify, Vercel deploy), invoke it from any agent. Includes a three-tier skill taxonomy, a reference executor, cryptographic skill signing, and a production skill set. |
Python (stdlib), Markdown, YAML, Supabase, Telegram Bot API, Vercel API | srmbsrg/ask-kernel |
| agent-spawner-demo | Dynamic agent instantiation with template inheritance, capability resolution, and full lifecycle management. Extracted from a multi-agent platform running concurrent domain agents — templates define blueprints, the spawner resolves inheritance at instantiation time, and a single command dispatcher handles all lifecycle transitions (spawn / start / pause / terminate / restart / assignTask). | TypeScript, Node.js | srmbsrg/agent-spawner-demo |
| multi-model-router | Per-stage multi-LLM routing with cost/quality/speed metadata, deterministic-first reasoning, and a hard token gate for budget enforcement. Extracted from a 4-stage code-generation pipeline (architecture → DB schema → API routes → UI components) that routes across multiple model providers — reasoning-heavy stages get a stronger model; latency-sensitive stages get a faster one. Daily token budget + per-minute rate limit cut runaway spend. | TypeScript, Node.js, OpenAI-compatible API | srmbsrg/multi-model-router |
| ThreadingTemplate | Async I/O patterns — concurrent task orchestration with thread pools and non-blocking operations. | C#, .NET, System.Threading, TPL, log4net | srmbsrg/ThreadingTemplate |
Perficient, Inc. — Senior Technical Consultant, App Modernization · Nov 2025 – Apr 2026 Delivered cloud-native .NET 10 microservices and enterprise API engineering for Lockton on a large-scale application-modernization engagement. Designed a Binder Management API suite using Clean Architecture and modular domain layers; integrated QuoteCompare AI to automate binder document extraction and premium calculation; implemented M-Files document-management integration; architected an Azure Redis distributed-caching strategy that reduced redundant Salesforce calls; built Salesforce integration microservices (Forza.Net / SOQL); and established CI/CD with NUnit/Moq test coverage.
Loomis Armored — Senior Software Engineer · Mar 2022 – Mar 2025 Led the modernization of Loomis Direct from legacy .NET to .NET Core and a React SPA, improving maintainability and UI responsiveness. Drove Azure cloud adoption and optimized CI/CD pipelines, cutting deployment time by 30%.
Transcore — Senior Software Engineer (Contract) · Oct 2019 – Nov 2021 Delivered modernization solutions for state-government systems using Angular, C#, and Azure-hosted SQL; converted multi-threaded legacy data processing into scalable, maintainable modules.
Roquemore — Senior Software Engineer · Jun 2015 – Sep 2019 Modernized internal WPF and MVC applications and led multiple third-party integrations including Equifax, TransUnion, and CarFax.
Automated Integration — Senior Software Engineer · Jan 2011 – Mar 2015 Revitalized a legacy ERP/dispatching system and developed C# / SQL Server field-service mobile applications.
- Backend: C#, .NET 10 / .NET Core, Clean Architecture, DDD, REST APIs, EF Core
- Front-end: Next.js, React, Angular, Blazor Server, React Native
- AI engineering: Multi-agent orchestration (Claude / Anthropic API), RAG & memory pipelines, agentic workflow design, prompt architecture, ElevenLabs voice
- Cloud & DevOps: Azure (App Services, Redis, Storage), Vercel, Railway, CI/CD (Azure DevOps, Jenkins, Git, Octopus)
- Integrations & payments: Salesforce (Forza.Net / SOQL), M-Files, Stripe, document processing
- Also: Node.js, Python, SQL Server, NUnit / Moq
Carborundum AI is an AI-native operations platform for small-to-mid trades businesses — the kind running on whiteboards and spreadsheets, underserved by enterprise ERP. The core components:
- Manifold — the ERP foundation: lead capture → customer record → sales order → dispatch → fulfillment → invoicing → payment, with a plugin/element architecture for vertical-specific add-ons.
- Tesa — a persistent AI operator running on the Claude API: inbound communications (SMS, Telegram, voice), RAG-based memory, and scheduled autonomous workflows.
- A multi-agent layer — AI officers operating with bounded autonomy under a written operating constitution, with persistent file-based memory across sessions.
The design philosophy: AI systems should be engineered, not prompted. Skills over prompts. Versioning over vibes. Composition over monoliths.
Email: scott@carborundum.ai · scottroymurphy@hotmail.com · scottroymurphy@outlook.com Web: scottroymurphy.com · carborundum.ai GitHub: github.com/srmbsrg LinkedIn: linkedin.com/in/scottroymurphy
Open to senior engineering, solutions-architect, and app-modernization roles where the work involves building real systems — .NET modernization, AI/agentic infrastructure, or both.