Documenting a daily journey through artificial intelligence — from first principles to deployed systems.
Hi, I'm Srimal Priyanga Fonseka — a software engineer (8+ years) now building toward an AI / Machine Learning Engineer career. AI Yathra ("yathra" = journey) is where I learn in public: each project is a checkpoint, built from the ground up rather than copied, so the reasoning is visible alongside the code.
📍 Belgium · open to relocation across the EU 🔗 LinkedIn · ✉️ hsrprf@gmail.com
| Project | What it demonstrates | Stack |
|---|---|---|
| Agentic Tool-Calling Demo | LLM-driven function calling, dynamic tool discovery, safe parameter parsing, and Python tool execution — a clean foundation for agentic workflows. | Python, LLM APIs |
| Pandas House-Price Pipeline | End-to-end ML data pipeline: cleaning, EDA, feature engineering, and a baseline regression model on the Kaggle House Prices dataset. | Python, Pandas, scikit-learn |
| XOR Neural Network | A multi-layer perceptron solving the classic non-linear XOR problem — built to show why hidden layers and non-linear activations matter. | Python, NumPy, Jupyter |
| Simple Brain (Perceptron from scratch) | Frank Rosenblatt's 1958 Perceptron implemented from scratch — the biological and mathematical roots of learning, via a "Pizza Detector". | Python |
🔭 Where this is heading: the next checkpoints in the journey move from foundations to a deployed end-to-end project (model + API + live demo on Hugging Face / Streamlit).
Languages: Python · SQL · JavaScript ML / AI: scikit-learn, neural networks (from scratch + frameworks), LLM tool-calling / agents Data: Pandas, NumPy, EDA, feature engineering Engineering foundation: software development lifecycle, system design, DevOps practices, REST APIs
Before AI, I spent years shipping web, mobile, and cloud software — as a software engineer, a startup co-founder, and a technology consultant for SMEs. That foundation in building and delivering real systems is what I'm now applying to machine learning.
📄 Full CV available on request · 💬 Best reached via LinkedIn