Hey, I'm Kamonphan — A Data Analytics Manager who builds end-to-end AI and automation systems, transforming complex machine learning problems into efficiency, measurable business outcomes.
01 · About
Data Scientist/ Analyst. Occasional instructor. Full-time bridge between "what does the model say?" and "what do we actually do about it?"
I didn't start in data. I started underground — literally. My geology degree at Chulalongkorn taught me to find signal in noise, read layers of complexity, and always ask "but what's actually going on here?" Turns out that's basically 80% of data science.
After a Master's in Data Science & Analytics at KMITL, I spent four years at VISAI doing everything from GenAI pipelines and LLM workflows to teaching machine learning concepts to rooms full of people who really didn't want to be there at 9am. (I like to think I won most of them over.)
These days I'm a Data Analytics Manager at Generali, where I lead fraud detection, cross-sell engines, and data strategy. My favourite part of the job is still the moment when a business stakeholder goes from "I don't really trust this model" to "wait, can we add more of these?"
Outside of work I write on Medium, break things in Python, and try to explain what I do to my family at dinner with varying success.
Timeline
Machine Learning & AI
Technical Stack
Management & Strategy
02 · Work
A few projects I'm proud of, plus articles where I try to make AI feel less intimidating — for engineers and non-engineers alike.
Insurance fraud is surprisingly creative. I designed a cross-functional detection system at Generali that spots high-risk health and accident claims — and built a daily automated alert so investigators aren't drowning in false positives.
Geologists spend a lot of time manually interpreting well logs. I built an ML pipeline to do the heavy lifting — cutting expert analysis time significantly and keeping the model honest with automated retraining and drift detection.
Predicting metal prices is hard. Building a system that stays accurate over time is harder. I designed a forecasting pipeline with automated retraining and data-drift monitoring so it doesn't quietly go wrong six months after launch.
03 · Contact
Got a project, a question, or just want to talk about AI and data? I'm pretty approachable — I promise.