
I identify the issue. I build. I review.
I am Sam Poplett, a software engineer and a Computer Science with AI student at Sussex.
The work
Each claim about a project names the file that proves it.
-
Provenance
A live globe of public open data
Road cameras, aircraft, earthquakes and more on one globe. Every layer names its publisher, its method and its limits. It is live, and the code is open.
A map layer with no stated limitation fails the build.
The file tests/
unit/ signals-explain.test.ts 92 stars and 15 forks on GitHub on 8 October 2026.
-

Unote
A notes app for university students
A block editor, flashcards, boards and a source library that checks each reference. It works with the network off. It is live on the web.
The source check asks a registry, not a language model. A source that it cannot reach is marked unreachable, never refuted.
The file server/
src/ lib/ references/ verify.ts
-

Kitchen Vision
A camera that follows people and knows them
The Pi only captures video and moves the servo. The laptop does all the machine learning.
The file head/
agent.py -

NewsScraper
World news, turned into events on a map
A quote that says where a person spoke to a reporter is not proof of where an event happened. The check rejects it.
The file newsfeed/
extract.py -

Prospector
A Reddit miner for unmet needs
A finding is dropped unless enough posts, subreddits and authors back it.
The file prospector/
report.py -

TrafficNerd
London traffic cameras in a terminal
The congestion score is the mean pixel difference between two frames of one camera.
The file trafficnerd/
image/ diff.py -

How I work
This is how I run a project with a client.
-

I identify the issue
Before I build, I speak to the client and identify the issue. I collect their first ideas, and I add my own where I can.
-

I build
I turn the client's vision into a product. Then I ask the client for feedback.
-

I review
I keep improving the product until the client is happy. Then I review it myself, so that it is reliable and easy to maintain.
-

I hand it over
The client gets instructions for the CMS, the documentation, and initial training on the tools.
Lots of this work is private, so these four steps name no file.
For my final review, I step back and look at the whole system.

At the desk
I am in the second year of Computer Science with AI at the University of Sussex.
I took client work in my first year, because the fast way to learn software engineering is to ship it for real people.
I write TypeScript and Python, from the database to the deploy.

Client work
Lots of my projects are non-disclosure bound, but I share what I can here.

-
Live data and analytics dashboard
Private project, United Kingdom
Recurring contract work. A dashboard that takes in data on a schedule, makes rolling backups, and shows trends over time. It is live in production.
-
AI call-handling platform
Private project, United States
Contract work. Phone calls in the browser, speech to text in real time, and a structured PDF report from each call.
-
Toughbook UI
An app made for a Toughbook. The user clicks words to type them, and AI finishes the sentences.
-
Tax organiser
A phone app for self-employed workers. It tracks tax, turns receipts into records, and exports the details to an accountant in a gold-standard format.
Tools in the pack
Each tool names one file that shows it in use.
-
TypeScript
Provenance, Unote
-
Python
five of the seven
-
React and Next.js
Provenance
-
Express and Postgres
Unote
-
FastAPI
Kitchen Vision, EasyBooks
-
SQLite
NewsScraper, Prospector
-
InsightFace on CUDA
Kitchen Vision
-
Electron
Unote
-
Vitest and Playwright
Provenance, Unote
-
pytest
the Python projects
-
MCP server
Prospector
-
systemd and CI
Provenance

Write to me
For a role or a project, write to me.
- Email lesttech.dev.sam@gmail.com
- GitHub github.com/011-sam-110


