Available for new projects

AI-Powered SaaS MVP Development

Build an AI-powered SaaS MVP designed for real business workflows, rather than one more chatbot. I help startups build SaaS products where AI can answer questions, automate repetitive tasks, and interact with your system securely from day one.

Experience building AI features with Semantic Kernel and MCP for production .NET SaaS platforms.

Let's build together → View portfolio →
Specialty

AI should solve real work

Many products add AI as a chat widget after launch. I design AI as part of the product from the beginning, so it can understand your business data, automate workflows, and grow with your application instead of becoming another disconnected feature.

  • AI that understands your business data, not a generic chatbot
  • Automation for the repetitive work your team does manually today
  • Secure AI access to your system from day one
  • An MVP delivered in weeks with a clear architecture
Start your AI MVP →
What you get
  • AI-ready SaaS architecture
  • Secure authentication & role-based access
  • AI assistants integrated with your data
  • Automation-ready backend
  • Multi-tenant SaaS foundation
  • Production deployment
  • Documentation & handover
What's included

AI Features I Can Build

AI customer supportDocument search & summarizationNatural language reportingInternal knowledge assistantsWorkflow automationEmail generationAI-powered dashboardsData analysisPDF generationAI agents using MCP
Good fit

Who This Is For

  • SaaS startups building an MVP
  • Existing SaaS products adding AI
  • Internal business tools
  • Workflow automation projects
  • Enterprise software modernization
Real work

Case Studies

MCP Server for Claude & ChatGPT

External AI agent access · US client

Built an MCP server on a live production platform so external AI agents, including Claude and ChatGPT, could query and act on real business data through one interface, authenticating as real logged-in users via OAuth 2.1/PKCE rather than a shared service account. One server, not two: the same MCP endpoint serves both agents. Implementing it surfaced several API design issues that would have affected developers and AI agents alike, fixed before new features were even built.

MCP ServersClaudeChatGPTOAuth 2.1 / OpenIddict

AI Built Into the Product

In-app AI assistant · US client

Built an in-app AI assistant with Semantic Kernel that lets users query live production data in natural language instead of navigating complex reports. Result: faster access to information, less manual searching, and an assistant wired directly into existing business data rather than a static knowledge base.

Semantic KernelLive-Data ChatEnterprise Reporting

MCP-First Development

A methodology born from that work

That experience shaped an approach I call MCP-first development: validate every API through its MCP interface, the same one AI agents use, before the UI is even built. AI checks the API against the real user story through that interface and writes the tests from what it finds, catching design gaps a UI would otherwise quietly work around. I use this on every new API I build now. Read how it works →

MCP-First DevelopmentAI-Assisted TestingAPI-First Design
Why work with me

I start by figuring out whether AI belongs in your MVP at all, and where it creates value, before touching the technology. When it does belong, I build it as a real part of your architecture rather than a bolted-on feature you'll have to rebuild later.

Process

How I Work

  • Understand your business before suggesting AI.
  • Decide whether AI belongs in v1 or later.
  • Build the core product first.
  • Integrate AI where it creates measurable value.
  • Ship AI features only after their output holds up against real data.
Toolbox

Stack & Technologies

Semantic KernelMCP ServersOpenIddictABP Framework.NET / C#ASP.NET CoreAngularBlazor.NET MAUIPostgreSQLRedisRabbitMQStripe
Common questions

FAQ

Should my MVP include AI on day one?

Not always. It depends on whether AI creates measurable value for your specific product yet; I'll say plainly whether it belongs in v1 or can wait, rather than adding it because it's expected.

What AI features deliver the fastest ROI?

Usually the ones that remove manual work your team already does. Document search, natural language reporting, and internal assistants tend to pay off faster than a customer-facing chatbot.

Can AI work with my existing database, or does it need to be greenfield?

It can work with an existing system. The MCP server work described above was built on top of a live production system, not a new one; the same approach applies to adding AI to what you already have.

Can you add AI later instead of building it into the MVP now?

Yes. If AI doesn't clearly belong in v1, I'll say so and design the architecture so it can be added cleanly later, rather than forcing it in early.

How do you keep AI features secure?

AI agents authenticate through the same OAuth 2.1/PKCE flow as real users, via OpenIddict, rather than a shared service account with broad access. Access follows the same role-based permissions as the rest of your system.

Not sure if AI belongs in your MVP yet?

Tell me what you're building. If AI creates real value there, I'll show you how to build it in properly; if it doesn't yet, I'll say that too.

Start a conversation →