AI vs Human Text Detection
Five deep learning architectures compared, BiLSTM, CNN, GCU, a hybrid model and a custom Transformer, for classifying AI-generated versus human-written text.
Best accuracy 98.4%
I turn data into decisions with machine learning, deep learning, NLP and modern data engineering, and I ship models as production services. Currently looking for junior Data Science, Data Engineering, AI, ML and DL roles in Finland.
I am a recent ICT engineering graduate from JAMK University of Applied Sciences, specialized in Data Analytics and AI, with a weighted grade average of 4.71 / 5.0 and top grades in Deep Learning, Machine Learning, Data Preprocessing and Computational Algorithms.
My work spans the full lifecycle: collecting and cleaning data, building and evaluating models, and shipping them as production services with FastAPI, Docker and Hugging Face Spaces. My bachelor's thesis delivered a deployed bilingual course similarity system used in a real university project.
Primary focus: Data Science, Data Engineering, AI, Machine Learning and Deep Learning. Secondary: full stack development with Node.js, Express, PostgreSQL and React.
Built for the LuoVuutta project: an evaluation of multilingual sentence embedding models for detecting similar course descriptions across Finnish and English in Finnish higher education. Five models were compared across 30 configurations. LaBSE performed best and now powers a production FastAPI service with a live public demo on Hugging Face Spaces.
Five deep learning architectures compared, BiLSTM, CNN, GCU, a hybrid model and a custom Transformer, for classifying AI-generated versus human-written text.
Best accuracy 98.4%
End-to-end ML on 3 million rows following CRISP-DM, deployed with FastAPI. Group project where I led business understanding, modeling and deployment.
R2 score 0.9456
LLM-powered tool that scores how well a CV matches a job description, using Llama 3.1 via the Groq API with PDF parsing and a Streamlit interface.
Llama 3.1 via Groq
Live interactive dashboard tracking the Finnish labour market with data from Statistics Finland, OECD and Eurostat, built with vanilla JavaScript and Chart.js.
Live, multi-source data
Machine learning on voice data, covering feature extraction from audio signals and model training for voice-based analysis tasks.
Audio feature pipeline
Full stack app with a secure REST API: Node.js, Express, PostgreSQL, JWT auth, node-cron recurring jobs and an external currency API, with a React and Vite frontend.
Live on Render
Graduated with a weighted average of 4.71 / 5.0 and 261 credits. Grade 5 (Excellent) in Deep Learning, Machine Learning, Data Preprocessing, Computational Algorithms, Data Analysis and Visualization, and the AI / DA project. Thesis grade 5.
Completed the full practical training module of the degree, applying data analytics and software engineering skills in real project work.
Supported daily IT operations, assisting with technical tasks, troubleshooting and customer-facing support across the company's services.
Advanced professional and academic English communication, with focus on grammar, vocabulary and writing.
Core skills in web development with HTML, Java programming and Microsoft Office tools for project work.
A practical introduction to backend programming. Learn to build a REST API with Node.js and Express, a core skill for modern web development.
Reflecting on the start of a new year as a chance for a fresh start: moving forward, growing and letting go.
More articles coming. Follow along at medium.com/@dggamage2000
I am actively looking for junior Data Science, Data Engineering, AI, ML and Deep Learning roles in Finland. If you are hiring or want to talk about a project, my inbox is open.