Case study · AI & data science
Preparing education data and forecasting student performance
For an education institute in Vietnam, we brought scattered data together, cleaned it and built models that predict student performance.
Project at a glance
- Client
- An education institute in Vietnam
- Period
- March to December 2019
- Team
- 4 people: 2 developers, 1 AI specialist, 1 project manager
- Technologies
- Python
- pandas
- scikit-learn
- TensorFlow
- Starting point
- The institute's data came from learning platforms, surveys and institutional reports, each in its own format, with duplicates and faulty entries. The institute wanted to use it for research and for curriculum planning, but in that state it could not analyse the sources together.
- Senva's role
- Senva collected the data, cleaned it, brought it into a common format and analysed it. On that basis we developed models that reveal patterns in the learning data and predict student performance from historical records.
- Approach
The data came first: removing duplicates and faulty entries, standardising formats, merging sources. Then came the analysis with pandas and scikit-learn, and the forecasting models with TensorFlow.
We worked in short iterations, because the institute's requirements changed as the project went on. The project was completed within the planned time and budget.
- Result
- The institute received a structured, analysed data set that let it base decisions on curricula and teaching methods on data.
Today's perspective
What we would document differently today
Under the EU AI Act, AI systems used in the EU to evaluate learning outcomes are classed as high-risk as a rule (Annex III, point 3, education). Under Regulation (EU) 2026/1744, the obligations for such systems apply from 2 December 2027. Today we would start a project like this with an EU AI Act classification and document the purpose, data and limitations of the models from the first day.
More case studies
- 2020-2023
Trade-fair platform for a telecommunications company
Interactive product presentation at trade fairs, with real-time analytics. React, Node.js, AWS.
Read the case study - 2018
Optical inspection of circuit boards
Several cameras and an ML model detect conductor tracks and components on circuit boards. Python.
Read the case study - 2017
Visual drone detection pilot
For Deutsche Telekom's drone shield programme. A CNN plus classical image recognition.
Read the case study
Intro call
Let's talk about your data
30 minutes with Kim Pham. We reply within one working day, and sign an NDA beforehand if you wish.
- info@senva.de
- Phone
- +49 30 5268 7541
- Address
- Bundesallee 32, 10717 Berlin, Germany