About

Hi, I'm Dan.

CS graduate, data scientist in training, lifeguard, and soon — full-time traveller. In June 2026 I'm leaving for a 3-year trip through the USA, Fiji, Australia, New Zealand, SE Asia and Canada. While I'm out there, I'll be running TravelNet — a personal data science project that collects and analyses everything from my GPS traces to my spending habits.

My academic background is in machine learning on time series data. My dissertation implemented and compared AutoPlait to existing time series segmentation algorithms for human activity recognition using the aeon Python library — work that feeds directly into TravelNet's ML pipeline.

I care about building things properly: reliable infrastructure, clean data pipelines, honest analysis. TravelNet isn't a weekend project — it's a 3-year longitudinal dataset I'm constructing from scratch, one commit at a time.

Dan Roberts

Canyonlands National Park during 4n4!

Data science & ML

My CS degree gave me the foundations; my dissertation gave me a real project to sink my teeth into. I'm interested in the messier, harder end of data science — real-world GPS traces rather than clean benchmark datasets, multi-currency financial data with missing values and format inconsistencies, health data subject to sensor dropouts. I enjoy looking for meaningful connections between data sources and using ML techniques to gain interesting insights, simply for the fact that they are interesting.

TravelNet is where that interest becomes practice. When I get to Australia I'll have three months of US data as a baseline and start the ML work: clustering travel behaviours, anomaly detection on spending, activity segmentation on GPS traces. I'll be able to automatically identify different workplaces I atteneded, coffee shops I visited, daytrips I took, activities I did and much more.

Python is my primary language. I use scikit-learn as my ML workhorse, with PyTorch on the table for anything that warrants it. I'm comfortable with Docker, SQLite, REST APIs, and running production systems on a Raspberry Pi from 10,000 miles away over Tailscale.

Lifeguarding & swimming

Before I was a programmer I was a swimmer. I've been on poolside for years; both swimming and guarding - water is something I (quite literally) cannot live without. This summer I'm returning to my favourite job: lifeguarding at a US summer camp for two months for the fourth year in a row.

It keeps me fit, keeps me outside, keeps me social, and is a good reminder that not everything worth doing involves a terminal window.

3 years, 9 countries

June 2026: I leave. The rough itinerary is Philadelphia/DC → USA summer camp → Seattle → Fiji (3-night stopover) → Australia on a Working Holiday Visa → New Zealand on a WHV → SE Asia backpacking (the Banana Pancake Trail, 3–4 months) → Canada on a WHV. Somewhere in there, a dissertation-worthy dataset gets built (and hopefully, some lifelong memories and friends).

The trip isn't a gap year. It's a deliberate decision to spend a few years being somewhere uncomfortable, meeting people I wouldn't otherwise meet, experiencing the world, and generating data I couldn't collect any other way. TravelNet is how I make that data useful.

2028-2029

Canada WHV

2027-2028

SE Asia — Banana Pancake Trail

2027

New Zealand WHV

2026-2027

Australia WHV — TravelNet ML begins

2026

Summer camp lifeguarding, USA

2026

TravelNet system built and hardened

2025

CS degree awarded, dissertation on time series ML