Projects
15 projects.
A bold website for a three-chair barbershop where customers pick a barber (or the first one free), book a time at midnight, and check the walk-in wait before they leave home. The owner runs the day from a calendar on his phone, and reminders go out on their own.
A warm, photo-led website for a coffee roaster with three cafés: the coffees with tasting notes and prices, subscriptions, brew guides, the story behind the roastery, and each café's hours. Fast on a phone, cheap to host, and with a support chat that answers from the business's own help center.
Order-ahead for a café with three locations. Customers order from their phone and pick a pickup time, orders appear on a tablet in the kitchen within seconds, and each table has a QR code for the menu. It also has a support chat.
A neighbourhood bakery's WordPress site with real bakery photography, a menu, custom cake orders with a recent cakes gallery, opening hours and a map, a blog, and a spot for ordering ahead. The owner edits it in the page builder she already knows, and it loads fast on a phone.
A chat widget that answers website visitors from a business's own help documents with the source, and books appointments from real open times in the business's booking system. One script tag installs it on any site.
A slow WordPress bakery site, fixed in place: home page from 47.0 MB to 456 KB, mobile Lighthouse from 62 to 96, a working order form, and visible on Google again. Same theme and editor, plus a support chat.
A calm website for a physiotherapy clinic. Patients book by appointment type and practitioner, new patients fill in their health form online before the first visit, and the answers are waiting on the booking when they arrive.
A friendly website for a small animal clinic. Owners book by pet and reason for the visit, the clinic keeps a record for each pet, and vaccination reminders go out on their own with a link that opens booking already filled in.
A cleaning company website where customers price their own clean in under a minute, or start from a photo of the space, with every line shown and tax separate, then send the quote in one click. Every request lands in a leads inbox the owner works from new to won to booked.
The private side of five small-business websites: a dashboard where each owner sees the day at a glance, keeps the calendar, customers and messages in one place, and changes prices, menus and hours without calling a developer.
Start a multi-step AI workflow and watch it run live: a separate worker runs each step from a job queue, retries what fails, and pauses for a person to approve, edit or reject the AI's draft before anything is sent.
Drop in a pile of invoices, receipts and forms, including scans and phone photos, and get back checked rows you can export to Excel or CSV. Every value shows where on the page it came from, and the math is checked before you approve it.
Type "waterproof women's hiking boots under $150, size 8" and the AI turns it into filter chips you can see, edit and remove. Every answer is checked against a schema, and an eval suite measures accuracy before any prompt or model change ships.
Full-stack work on the core search of a marketplace used by millions of students: a React and TypeScript catalog with 150k+ listings and 50+ filters, saved searches and compare, a NestJS API layer over OpenSearch, and natural-language filters.
The site you're reading. Statically generated Next.js, content edited in a CMS admin, and a full technical SEO layer.
Papers from my MSc research on bin packing, published at computer science conferences, plus earlier data science projects and the startup I founded.
Research paper2020
A research paper on square packing, a two-dimensional version of bin packing where square items go into as few square bins as possible. Unlike earlier models, items may be rotated, which matches how stock cutting works in practice.
Research paper2021
A research paper on placing replicated tenants onto as few cloud servers as possible while staying safe when a server fails. Tenants arrive one at a time and can't be moved later, so this is an online bin packing problem with fault tolerance built in.
Research paper
A desktop application that takes a developer's question, collects Stack Overflow results in real time, then ranks and clusters the code snippets in the answers so the useful ones are quick to find.
Data science2021
Classification models that predict the intention behind a code review question, such as asking for a clarification, suggesting an improvement, or asking why something was built a certain way, using the question's words plus key-question, sentiment, and length features.
Data science2020
Visualizations and regression forecasts of the early COVID-19 outbreak, from 22 January to 7 April 2020, using Johns Hopkins CSSE data, comparing polynomial, Bayesian ridge, and support vector regression.
Data science2014
Data mining on 6,000 abdominal aortic aneurysm patient records to find how medical records relate to length of stay at the hospital. After careful preprocessing, some models reached 70 to 84 percent accuracy.
Product2012
A social network connecting local businesses and their customers. It started as a business plan that was named the "best promising project" at the IUT Entrepreneurship Festival, and I then founded a startup to build it.
Send a short note on what you're building. I'll reply with questions and a rough plan.
