Profile

Software developer specializing in AI engineering. I build production software and agentic AI systems that connect LLMs with enterprise data and real-world workflows.

My current focus is Microsoft Fabric, Microsoft Foundry, RAG and multi-agent architectures. Behind that is nearly four years of full-stack development across backend, frontend, infrastructure and production systems.

Experience

Software Developer

Insta Advance Oy · 03.2026 – present

  • Built a multi-agent proof of concept for factory OEE analysis: an orchestrator agent coordinating a Microsoft Fabric data agent (RAG) and two specialized Microsoft Foundry agents.
  • Agents used external tool calls to calculate OEE from time series data and to compare production line data against work shifts to identify discrepancies.
  • Developed further agentic AI proofs of concept integrating enterprise data with modern LLMs.

Junior Software Developer

Insta Advance Oy · 08.2023 – 03.2026

  • Designed and built a completely new version of the time series data management UI in Insta Industrial Dataplatform (Node.js, TypeScript, Angular).
  • Continuously developed a certificate lifecycle management solution in Insta PKI Solutions (Go, React).
  • Re-created an internal PHP time-saving tool in Go.
  • Acted as Scrum Master for a five-person team in 2025.
  • Maintained production Linux servers as needed.

Software Developer Trainee

Insta Advance Oy · 02.2023 – 08.2023

  • Modernized dependencies and code of a Node.js and Vue.js application.
  • Hardened CentOS 7 production servers and built backend features.

Projects

Little Closet

Full-stack side project · Go, Next.js, PostgreSQL, Microsoft Foundry

A multi-user app for organizing baby clothing by size and category, with AI that recognizes items from photos and recommends outfits for the weather.

  • Image analysis: a photo is analyzed by an AI model that returns the item name, category, size and whether the size came from the tag or was estimated. Results are matched to the user's own categories and sizes, then pre-fill the form for review.
  • Outfit recommendations: a serverless function reads the user's wardrobe, fetches live weather from Open-Meteo and asks a Microsoft Foundry model for an outfit, in Finnish or English.
  • Secure by design: Google sign-in through Neon Auth, with a Go API that verifies the JWT on every request and scopes all data to the signed-in user.
  • Self-hosted: the frontend and API run in Docker Compose behind Caddy on my own Linux VPS. The app is an installable, offline-capable PWA.
Little Closet architecture A browser PWA reaches Caddy on a Linux VPS, which routes to a Next.js frontend and a Go API. The Go API uses Neon Auth, a Neon DB Lakebase Postgres database and Neon Functions, which call Microsoft Foundry and Open-Meteo. Self-managed Linux VPS · Docker Compose Browser installable PWA Caddy HTTPS reverse proxy Next.js React frontend Go API JWT-verified Neon Auth Google sign-in, JWT Neon DB Lakebase Postgres Neon Functions serverless AI Microsoft Foundry image analysis, outfits Open-Meteo live weather
How Little Closet fits together.

Open the app · View source on GitHub

Skills

AI engineering
Agentic and multi-agent solutions, RAG, LLM integration, tool calling
Languages
Go, TypeScript, Python
Frameworks
Node.js, React, Next.js, Angular, Vue.js
Data and cloud
Microsoft Fabric, Microsoft Foundry, Azure, PostgreSQL, time series data
Infrastructure
Linux server administration, Docker Compose, Caddy
Ways of working
Scrum, agile, team collaboration

Certifications

  • Microsoft Certified: Azure AI Engineer Associate
  • Microsoft Certified: Azure Fundamentals
  • Foundation Certificate in Project Management
  • CCNAv7: Introduction to Networks

Education

Business Information Systems, BBA
Tampereen ammattikorkeakoulu · 2020 – 2023