Data Scientist · ML Engineer · Co-founder, Kinnovation
I started in psychology because I wanted to know why people do what they do. Four years in, I realised the answers I wanted were sitting inside datasets nobody had bothered to build properly. So I went and learned that too: a postgraduate diploma at UT Austin, then the DACSS master's at UMass Amherst. I am not the person who stops when something works. I am the person who asks what it would take to make it right, and then goes and does that part as well.
Entrepreneurs Accelerator Program · Greenfield, MA
Selected for the Spring 2026 cohort. Ten weeks of having my assumptions taken apart.
Data & Behavioural Insights Associate · Menifee, CA
A year building the analytics for a legal practice, and rebuilding the site that fed it.
Client Experience & Digital Strategy Consultant · Remote
A wellness practice that was good at the work and bad at explaining it. I fixed the explaining.
Behavioural Health Technician · San Diego, CA
Applied Behaviour Analysis with autistic children. The hardest job I have had, and the one that shaped everything after.
Featured Speaker, AI for Mental Health · Blacksburg, VA
Invited to speak on machine learning for personalised mental health intervention.
Data Analytics Intern · Doha, Qatar
Six months building datasets from scratch and the automated reporting that ran on them.
Data Analytics Intern · Doha, Qatar
Designed a mobile data plan from how customers actually used their phones.
Student Manager · Blacksburg, VA
Five and a half years running shifts in a high-volume dining hall, through the whole degree and the diploma.
UMass Amherst
I pulled 6,398 posts and 19,488 comments out of three mental health subreddits and put three classifiers against each other on the sentiment labels. Every file is in the repo.
Python · PRAW · scikit-learn · NLTK VADER · Transformers · Plotly Dash
UMass Amherst, DACSS 758
A scraper that builds one comparable dataset out of ten universities that all publish their crime alerts differently. 519 documents, HTML and PDF, with full provenance on every row.
Python · BeautifulSoup · pdfminer.six · cloudscraper · requests
UMass Amherst, DACSS 690S
A D3 and Three.js scrollytelling piece on how pop music changed across 24 years, built on Billboard chart data joined to Spotify audio features.
D3.js · Three.js · JavaScript · Python
UMass Amherst
An adversarial search agent for misere Nim. Iterative-deepening minimax, alpha-beta pruning, transposition caching, and a domain evaluator, all inside a one second per move limit.
Python · Minimax · Alpha-beta pruning · Memoisation