About
From ship decks to data pipelines.
I spent eight years at Silja Line leading cabin service teams on passenger ships. It was a job about people, processes and keeping things running under pressure.
In 2021 I started studying computer applications at HAMK. My thesis combined parking-space detection with a chatbot, and an internship at HAMK Smart had me building computer vision for industrial quality inspection. Before that, a project with Cinia had me forecasting call volumes from weather and calendar data.
I'm now doing an M.Sc. in Computer Science (Machine Learning & AI) at the University of Helsinki. Most of my time goes into ML systems on real data and into how LLM-based systems fail. On the side I run Barmo, a small business I started in 2026 for websites, AI automation and bookkeeping, and I build apps for the Nepali community.
How I work
Baseline first
A model is only as good as what it beats. The tram forecaster is measured against persistence, and FinTax measures retrieval before anything is generated.
Real data, real problems
Live transit feeds, ten years of public procurement records, atmospheric measurements and genome data. Messy data is where most of the work, and most of the learning, happens.
Finish the whole loop
Tests, CI, a deployed dashboard and a write-up of what didn't work. A notebook result is the middle of a project, not the end.
Experience
2026 – present
Founder (side business) · Barmo
- Built barmo.fi, a bilingual site prerendered to static HTML.
- Built Barmo Accounts, a bookkeeping service with an OCR + local-LLM pipeline that pre-fills documents for human review.
- Runs BPIL, an R&D project on Finnish public procurement data.
Jun 2024 – Nov 2024
Machine Learning Engineer Intern · HAMK Smart
- Compared a classical machine-vision approach (OpenCV) with deep-learning detection (YOLO).
- Built a Django web app for inspecting samples and comparing them against standard reference images.
- Worked on model optimisation and inference speed.
Sep 2023 – Jan 2024
Summer Project Developer · Cinia Oy
- Combined call history with Finnish weather data and calendar events as external factors.
- Built the data pipeline and ARIMA / SARIMAX forecasting models.
- Created dashboards for exploring trends.
Jul 2013 – Aug 2021
Service Manager · Silja Line
- Coordinated staff and daily operations in a high-pressure, customer-facing environment.
Education
2025 – present
University of Helsinki
M.Sc. Computer Science — Machine Learning & AI
Coursework and research in machine learning, multi-agent LLM systems and data science.
2021 – 2024
HAMK University of Applied Sciences
Bachelor of Computer Applications
Thesis: Smart Parking Assistant (YOLOv5 + Rasa chatbot).
Skills
- Machine learning
- PyTorch · scikit-learn · Gradient boosting · Computer vision (YOLO, OpenCV) · NLP (Hugging Face, spaCy) · Time-series forecasting
- LLMs & retrieval
- RAG pipelines · LangChain / LangGraph · Ollama (local models) · ChromaDB / Qdrant · Evaluation of LLM systems
- Data & MLOps
- Pandas / Polars · DuckDB / Parquet / SQL · Entity resolution · MQTT streaming · Docker · GitHub Actions CI · pytest
- Software
- Python · TypeScript · Next.js / React · FastAPI / Flask / Django · PostgreSQL / Supabase · Electron
Certifications & languages
AWS Academy Cloud Foundations
Amazon Web Services · 2023
Data Analyst in Power BI
DataCamp · 2023
- EnglishFluent
- FinnishIntermediate
- NepaliNative
Want the details on a specific project? Get in touch.