The Journey of a Faceless Robot: Why Data Scientists Should Learn a Little Web Development
Imagine building a robot. It’s powerful.
It can analyse data, detect patterns, and make intelligent predictions.
It does exactly what it was designed to do but there’s one problem.
It has no face.
No interface.
No way to interact with the outside world.
No way for people to understand what it actually does.
So, when someone looks at it, they ask:
“What am I even looking at?”
Well, that robot… was my data science work.
Stage 1 - The Robot in The Lab
In environments like LASER – Trusted Research Environments, and across data science more broadly, most of our work happens behind the scenes.
We focus on cleaning the data, building statistical methods, data analytics, building machine learning models, deep learning models, optimizing performance. And in more recent years, this also includes working with artificial intelligence (AI), from predictive models to more advanced approaches.
But it stays in the lab, hidden in notebooks, buried in code, visible only to those who already understand it and a mystery to people with no coding or prior knowledge.
Stage 2 – Understanding… But Not Quite Seeing
At the start of my data science career, I found myself in a strange position.
I understood Why I was doing the work, I understood How to do it (to some extent). I could follow the logic, but I couldn’t fully see it in action.
It felt like I had built something important… But I couldn’t quite grasp how it lived in the real world.
Stage 3: The Question That Changed Everything
Then came the big one, my Masters dissertation.
For my project, I worked on developing a deep learning model to detect signs of leukaemia from blood smear images. In simple terms, the model takes an image of a blood sample, analyses patterns in the cells, and predicts whether signs of leukaemia are present.
Technically, it worked.
But I realised I didn’t want to stop at just building the robot as usual.
I kept asking myself: how does this actually connect? How does my work reach people, become usable, and make sense to someone who isn’t technical?
Everything I had built lived inside my notebook. To explain it, I had to walk through code, describe the model, and interpret the results step by step. It made sense to me, but I could see how quickly it became unclear to anyone else.
There had to be a better way to show what my work actually does. Something more accessible. Something that didn’t just explain the work but demonstrated it.
Stage 4: The Robot Gets a Face
That’s when I discovered flask and everything changed.
Flask is a lightweight Python web framework that allows you to take your code and expose it through a simple web interface. In practice, that means you can connect your model to something a user can interact with, without needing to build a full-scale web application.
For me, it became the bridge between:
- Backend models
- Real-world interaction
It didn’t require deep frontend knowledge, it didn’t overcomplicate things, it just gave me a simple way to make my work interactive.
That shift from something that only made sense in a notebook to something that made sense in context was small in effort but huge in impact.
And just like that…
The robot got a face.
Stage 5: Transformation.
Before Flask, my work looked like Code in notebooks, outputs in cells, numbers and percentages. After adding an interface, it became something else entirely.
It was nothing fancy, as you can see in the image above; just a box to upload an image, a button that said, ‘predict image’, but now you could interact with it.
Instead of explaining what the model does in jargons that leave people confused you could demonstrate in real time what you’ve done.
Input -> Action -> Result.
In my case, upload an image, click a button, see the result instantly.
Why This Matters for Data Professionals
A model sitting in a notebook is technically complete, but practically limited. The moment you place it behind a simple interface, it becomes something people can interact with, not just something you explain.
That changes more than just usability, it changes outcomes.
When stakeholders can clearly see and interact with your work:
- They can make faster and more confident decisions
- They can give more meaningful feedback
- They better understand the impact of the model
- They’re more likely to trust and reuse your work
It also makes collaboration easier. Instead of long explanations, you’re showing rather than telling making it easier for non-technical teams to engage with what you’ve built.
More than anything, it forces a shift in thinking. You stop focusing only on how the model works and start thinking about how it’s used, what the user sees, what they need, and what they take away. That perspective connects the technical side of your work to its real-world purpose.
And in a field where many people stop at analysis, being able to take that extra step even in a simple way can set you apart.
A Quick Reality Check
Of course, none of this replaces the core work. The data still needs to be clean, the models still need to be sound, and the analysis still needs to be meaningful.
But without a way to interact with it, even good work can remain underutilised.
THE END
By Obosekokhune Eselebor
Research Software Technician
LIDA Data Analytics Team
Read more blogs from Obose Wrangling Chaos: 6 Things I Wish I Knew Before Tackling Messy Data
The Data Analytics Team are the in-house, technical LASER support at LIDA
The team provide end-to-end support from project conception to closure, and design personalised data environments tailored to your data and requirements. They are a team of dedicated specialists in data management, data analysis, software engineering and information governance, who collaborate with LIDA researchers across all stages of their projects.