Back to our work
Applied ML

Predicting learning outcomes for around 100 institutions

A service that trains a model on each institution’s data, explains its predictions and analyses text, topics and sentiment. It launched in 2021 and was extended in 2025-2026.

Period
Jun 2021 - Feb 2022 · Modernised 2025-2026
Our scope
ML service & API
Engagement
Delivered by the aiLabs founders
One prediction model per institution An institution’s data is used to train neural network and logistic regression candidates. The selected model is available through the prediction API. One model per institution Institution data Neural network Logistic regression Selected model Prediction API
Explanatory diagram illustrating the engineering approach.

Key results

  • Used by around 100 educational institutions, most of them in the United States.
  • In production since 2021.
  • Predictions, data insights and text analytics through one API, without a data scientist for each customer.
  • Modernised and extended with new functionality in 2025-2026 during a second engagement with IntelliBoard.

The challenge

IntelliBoard wanted to give educational institutions predictions about learning outcomes, such as which students are at risk of not completing a course. Every institution has its own data: often a few hundred records with 5-20 attributes, and far more students who complete a course than who do not.

Our contribution

  • Designed a Go API and a queue-based architecture in which training, prediction, data analysis and text analysis run as separate Python services.
  • Trained a separate model for each customer dataset: a compact PyTorch neural network together with an interpretable logistic regression, with SMOTE oversampling for imbalanced classes.
  • Trained several candidates per dataset and kept the one with the fewest prediction errors. Model weights, ROC curves, detailed error statistics and a data report are stored in PostgreSQL.
  • Estimated training time from the history of similar tasks, so the product could tell customers when a model would be ready.
  • Built automated summaries highlighting unusual values, correlations and significant attributes, and text analysis: language detection for 12 languages, readability, topics and sentiment.
  • During the second engagement in 2025-2026, moved the services from RabbitMQ to Kafka, introduced managed database migrations and added linear regression with model coefficients.

What the work delivered

The multi-tenant prediction service went into production in 2021 and was used by around 100 educational institutions. aiLabs returned in 2025-2026 to modernise the service and extend its predictive capabilities.

More engineering stories

View all work
Data engineering

Enterprise data warehouse

A BigQuery warehouse that brought sales, e-commerce, web analytics and sales-force data together, with Power BI reporting and market basket analysis for the US and Canada.

Applied ML

Real-time bidding optimisation

A bid optimisation system for advertising campaigns on The Trade Desk: event data pipelines, simulation-tested algorithms and services that adjust bids across many dimensions every hour.

Computer vision

Planogram compliance

As part of the client’s team, we built the 2D matching engine and backend. Cutting matching time for 80 products on one side of a fixture from 53 to 3-4 seconds made photo-based planogram checks practical on the shop floor.

Let’s build something useful

Bring us the
hard part.

A new AI capability, a complex data problem or a platform that needs to evolve. Tell us what you’re working on.

Discuss your project

Start a conversation

info@ai-labs.tech

aiLabs Innovations FZ-LLC
Compass Building, Al Shohada Road, Al Hamra Industrial Zone-FZ, Ras Al Khaimah
United Arab Emirates
Licence No. 47014246

For an initial conversation, share your goal, the current system and the challenge you want to solve.