Paideia Mundi Neural Research Platform

The Problem

Early cognitive screening in young children relies heavily on trained observers, which is slow, costly and varies between observers. The institute wanted to explore whether a short, non-invasive brain recording could add a complementary, more objective signal for clinicians and researchers.

My Role

AI/ML Engineer - Team Lead of a small cross-university team (March to May 2026), reporting to the institute's leadership. Owned scope, standups, stakeholder communication and end-to-end delivery, plus data preprocessing, machine learning, pipelines and the dashboard.

The Approach

• Built ETL pipelines for EEG and proprietary datasets, turning raw recordings into clean, model-ready signals and engineered features.
• Compared several model families side by side with uncertainty estimates and significance testing, so recommendations rest on evidence.
• Shipped a sign-in protected dashboard that explains predictions and model behaviour to clinicians and researchers without an engineering background.

Key Decisions

• Prioritised honest, defensible findings over impressive-looking ones, so the institute could trust and build on the results.
• Benchmarked multiple approaches before committing, rather than picking the one the team knew best.
• Designed the dashboard and documentation for non-technical leadership first, because the work only matters if the people steering the research can use it.

Impact & Results

• Delivered the research prototype end to end within the program timeline: pipeline, models, dashboard and documentation.
• Kept quality high with automated tests and continuous integration.
• Co-presented the live demo to the institute's leadership and handed over full technical and plain-language documentation.

The Stack

What I Learned

• Leading a team across universities is mostly about clear scope and fast feedback.
• In research, a trustworthy result beats an impressive fragile one.
• Writing for clinicians changed how I built the product: every view needed a plain-language explanation.