I'm a software engineer with 10 years of commercial experience across e-commerce, energy, and enterprise domains. My core stack is .NET, C#, Azure, and microservices - building distributed systems that handle real load on real infrastructure.
When I build something useful and repeatable, I open-source it. Right now that means AI observability platforms, multi-agent developer tooling, and educational .NET content.
I also run Dev Insight — a channel covering C#, .NET, and system design through deep dives, flashcards, and live coding.
Below are the open source projects I actively maintain, focused on AI tooling, developer experience, and educational .NET content.
| Project | Description |
|---|---|
| architecture-standards | An architecture constitution for .NET Aspire on Fly.io and Azure - 15 principles and 17 operational guides, readable as documentation and installable as agent plugins for Claude Code, Copilot and VS Code |
| agent-eval-bench | A spec-first evaluation bench for tool-using agents: 35 YAML scenarios, 313 assertions over the execution trace, a deterministic layer that gates every pull request, and a rubric judge that reports skipped rather than green without credentials |
| CopilotScope | AI coding-session observability - an OTLP collector with a hand-written protobuf decoder, a Postgres store, and a Blazor dashboard that scores session quality rather than token spend, across five assistants |
| AgentHelm | Web cockpit for AI coding agents - drive GitHub Copilot CLI, Claude Code, and Gemini via the Agent Client Protocol with permission policies, audit trail, and git diff review |
| authservice | Standalone auth microservice for ASP.NET Core: JWT with rotating refresh tokens and reuse detection, RS256 with published JWKS, Google/GitHub OAuth with provider-side email verification, multi-tenant organizations, and an append-only audit log |
| MAF for .NET Engineers | Practitioner's guide to Microsoft Agent Framework 1.0 for .NET - agents, tools, MCP, graph-based workflows, multi-agent orchestration, middleware, and hosting, with samples that run against a local model |
| LangChain, LangGraph and Async Python | Practitioner's guide to LangChain 1.x and LangGraph for engineers arriving from .NET - async Python, cancellation, persistence, interrupts, context engineering, and five specified measurement experiments |
| csharp-flashcards | Beamer Q&A flashcard deck covering C# from fundamentals through cloud and leadership topics - built for interview prep, classes, and self-study |
| agents-and-llms | Searchable knowledge base for the Agents & LLMs shorts series - static, search-first page covering basics, foundations, model internals, and agent tooling, with content driven entirely from JSON |
| bayesian-inference | A from-scratch exact Bayesian-network inference engine (Enumeration-Ask, no probabilistic-reasoning library) in C#, wrapped in a Blazor WebAssembly and ASP.NET Core app |
| black-hole-sim | A Schwarzschild black hole raytracer in C#: numerically integrates photon geodesics via RK4 and renders a thin accretion disk, with a console renderer and an ASP.NET Core API |
| TinyTransformer | Minimal educational C# implementation of the Transformer architecture - encoder-decoder, multi-head attention, feed-forward layers, and positional encodings |
| llm-multi-agent-chess | Two LLM-powered chess agents battle through a Streamlit interface with configurable personalities and real-time board visualisation |
| chess-mas | Multi-agent chess in C#, orchestrated with .NET Aspire and Microsoft Agent Framework, with cost tracking against Azure AI Foundry |
| IberiaFamilyCalculator | Interactive Spanish IRPF, autónomos, and payroll calculator - drag sliders, add entities, share the full setup via a single URL |
| dev-insight | Practical coding examples and patterns featured on the Dev Insight channel - C#, .NET, Transformers, SOLID, and more |
| C# Dictionaries - Deep Dive | Deep dive into Dictionary internals - hashing, buckets, collisions, resizing, complexity analysis, and benchmarks |
This repository also hosts a data-driven portfolio map: thirty-four repositories, bilingual,
with a consolidation analysis and an operating model behind them, rendered as a small
static site. See docs/PORTFOLIO-MAP.md for what it is, how it's
maintained and how it deploys.
Mostly AI tooling, educational .NET content, and long-lived open source platforms rather than throwaway demos.