Fransiscus Setiawan

AI Solution Architect · Sydney

AI solution architect Sydney practice: Azure, .NET, agents & RAG, and enterprise integration — with multi-tenant ACL and evals before cutover.

Azure · .NET · Agents & RAG · Enterprise integration

Who is Fransiscus Setiawan? Fransiscus Setiawan is an AI Solution Architect based in Sydney, Australia. He builds AI agents, MCP servers, and enterprise AI systems using .NET, Azure, and production software architecture practices.

What does Fransiscus Setiawan build? He designs and implements systems that connect language models with real business workflows: retrieval pipelines, tool-calling agents, Azure AI integrations, event-driven services, and secure APIs.

AI Engineering Focus

My current work is centred on practical AI engineering rather than demos: MCP servers, agent orchestration, RAG systems, evaluation loops, local model experiments, and cloud deployments that can be operated by a real engineering team.

Enterprise AI on Azure and .NET

My background in .NET, Azure, and distributed systems helps me design AI products that respect integration boundaries, reliability requirements, data security, and deployment constraints. Typical building blocks include Azure OpenAI, Azure AI Search, Azure Functions, Durable Functions, Service Bus, ASP.NET Core, and SQL-backed services.

EV Charging and Smart Mobility

I have worked deeply in EV charging platforms, including OCPP-based systems, charging network integrations, and operational architecture. That domain experience informs AI work around diagnostics, anomaly detection, load-management assistance, and support workflows for charging infrastructure.

Technical Background

  • AI: MCP, tool calling, RAG, Claude API, Azure OpenAI, Azure AI Search, local LLM experimentation, evaluation workflows
  • Engineering: C#, .NET, Python, TypeScript, ASP.NET Core, REST, GraphQL, event-driven architecture
  • Cloud: Azure Functions, Durable Functions, Service Bus, Key Vault, Azure DevOps, GitHub Actions, Docker
  • Domains: EV charging, OCPP 1.6/2.0.1, enterprise integration, SaaS platforms, fintech, retail, energy

Professional Profile

I have worked as a Lead Engineer, Engineering Manager, Head of Engineering, and Solution Architect across enterprise software teams. This site is my technical archive: articles, architecture notes, experiments, and case studies from building production systems.

Code and open notes: GitHub. For professional contact, use the contact page.

FAQ

What does an AI Solution Architect build?

Production AI systems that connect models to real workflows: agents with tool/MCP boundaries, RAG with citations, Azure OpenAI integrations, and APIs that ops teams can run. Not demos — contracts, evals, and failure modes. See model-agnostic AI architecture.

What stack does Fransiscus use?

Azure and .NET as the default spine: Azure OpenAI, Azure AI Search, Functions / Durable Functions, Service Bus, ASP.NET Core, plus TypeScript and Python where they fit. Focus areas: agents & RAG, MCP gateways, and enterprise integration.

How do you handle multi-tenant ACL on RAG?

Filter by tenant and user rights in retrieval before chunks enter the prompt. The model is not the ACL layer. Search filters narrow documents; Entra (or your IdP) still authenticates the caller. Related: security-review decisions.

How is an AI Solution Architect different from an AI engineer?

An AI engineer often owns model/feature delivery. An AI Solution Architect owns the system boundary: contracts, tool policy, retrieval ACL, eval gates, observability, and how the work lands on Azure/.NET integration. Both ship code; the architect role is accountable for what still works after go-live.

How do you evaluate agents and MCP tools in production?

Golden sets with pass/fail (schema, tools, citations, refuse), regression vs the current deployment, and canary with a kill switch — before write tools stay open. Details: evaluating AI agents in production.