Write-up · EdTech
HEXDigital Content Library Builder
Short, hands-on course content for students heading into work, and a way to see whether learners actually make it through.
Content and research at HEX, an education startup, November 2023 to June 2024
HEX is an education startup, and its courses, platform and learner records belong to HEX. This page is my own write-up of the work. The demo further down is a concept illustration that I wrote fresh for this page, with a made-up module and synthetic learners. It is not HEX's platform, content or data.
- Period
- Nov 2023 – Jun 2024
- Students supported
- 80+
- Area
- EdTech startup
- My part
- Content, research, prototypes
The problem
Many university students worry that a degree on its own will not be enough to land a first job. Employers also look for practical skills and some sense of how work actually happens. HEX is an education startup that works on that gap, with short courses and programs that sit alongside a degree.
Short online courses have a well-known weakness. People start them with good intentions and then drift away part of the way through, and the team running the course often cannot see where, or why. Content that asks learners to do something, rather than only read, tends to hold attention for longer. It only helps, though, if you can measure whether it worked.
What I did
Over about eight months I worked on four connected pieces of the same question, which was how to keep learners engaged and how to tell whether it is working.
Interactive content
I built and refined interactive digital content for HEX's course library, aimed at helping students move from study into work.
Research
I researched how newer technology, including large language models, could make courses more engaging and improve what learners get out of them.
A catalogue prototype
I built a front-end prototype of a page where students could browse short courses, search and filter them, and collect the ones they wanted in a cart.
Engagement analysis
I started an analysis pipeline for learner engagement, planned around time spent on each module, how scores were spread and where learners dropped off.
How it is built
- Next.js
- React
- TypeScript
- NextUI
- Tailwind CSS
- Zustand
- Python
- pandas
- Plotly
- Jupyter
- Notion
- LLMs
The catalogue prototype used the Next.js App Router with NextUI components and Tailwind CSS, written in TypeScript. It had a course listing with search, filters and sorting, and a small cart kept in the browser's session storage, so a student could gather courses before checking out.
The analysis side was a Python environment with Jupyter notebooks, pandas and Plotly. I set it up so the same steps could later run on exports from more than one course platform and be compared. It was still at an early stage when my time at HEX ended, so this page shows where it was heading rather than any results.
The demo below is new. I wrote it in JavaScript and React for this site, and it shares no code with anything at HEX. Its learners come from a seeded random number generator, so the numbers stay the same on every visit until you ask for a new cohort.
Try a module, then see the cohort
Concept illustration on synthetic data. The module is one I wrote for this page, and the learners on the dashboard are generated in your browser. None of it comes from HEX's courses, platform or learners, and it is not the real system.
The first panel is a sample micro-course module with a concept card, a worked example and a three-question quiz that gives feedback straight away. The second panel is the kind of engagement dashboard my analysis was heading towards. When you finish the quiz, your score and your time join the dashboard as You. Nothing you do here is stored or sent.
Learner view
Module 1 of 4
Problem statements
One sentence, four parts
A good problem statement fits in one sentence. It names who is stuck, what gets in their way, why it matters and how you would know it is fixed.
Leave the solution out for now. A solution in the sentence closes off better ideas before you have had a chance to look for them.
Who
A specific group, not everyone.
What gets in the way
The moment or obstacle where they get stuck.
Why it matters
What it costs them, or the people around them.
How you would know
A sign you could actually measure.
Engagement dashboard
Synthetic data · 120 learners · seed 2024
- Learners
- 120
- enrolled
- Finished the course
- 60%
- 72 of 120
- Biggest drop
- M2
- 20% stop in Module 2
- Average score
- 2.2
- out of 3, all quizzes
Module 1 · Problem statements
Show the numbers
| Stage | Learners | Share |
|---|---|---|
| Enrolled | 120 | 100% |
| Module 1 | 103 | 86% |
| Module 2 | 82 | 68% |
| Module 3 | 74 | 62% |
| Module 4 | 72 | 60% |
| Module | 25th | Median | 75th |
|---|---|---|---|
| Problem statements | 4.6 | 6.1 | 8.2 |
| Talking to users | 7.8 | 10.0 | 14.2 |
| Quick prototypes | 11.0 | 14.0 | 18.9 |
| The short pitch | 6.0 | 7.8 | 9.6 |
| Module | 0 | 1 | 2 | 3 |
|---|---|---|---|---|
| Problem statements | 2 | 18 | 38 | 45 |
| Talking to users | 9 | 16 | 33 | 24 |
| Quick prototypes | 0 | 12 | 27 | 35 |
| The short pitch | 0 | 11 | 32 | 29 |
Every learner here is generated in your browser by a seeded random number generator. Change the seed and the cohort changes with it. Nothing you do on this page is stored or sent.
My role and what I learned
I worked at HEX from November 2023 to June 2024, on content, research, the catalogue prototype and the engagement analysis. It was my first role at a startup. The content I worked on supported more than 80 students.
Most of what I took away was about measurement. It is easy to make content feel more engaging, and much harder to show that it changed what learners did. Asking how we would know it worked, before building anything, is a habit I carried from HEX into later work. It is also the last question in the quiz above.
HEX's own site lists its current courses and programs.