Mô tả công việc
Analytical Research & Modelling
Employ dimensionality reduction (PCA, UMAP, t- SNE) and unsupervised learning (clustering, mixture models) to discover latent learning traits.
Conduct exploratory data analysis (EDA) on handwriting and voice datasets to identify behavioral patterns and anomalies.
Build and interpret regression models — linear, logistic, mixed effects, LASSO, Ridge — to isolate key factors influencing performance, engagement, or stress.
Perform feature engineering from raw handwriting and audio data (e.g., hesitation index, cognitive delay markers, pitch variability).
Apply causal inference frameworks — backdoor criterion, DAGs, propensity scoring, mediation analysis — to uncover genuine cause- effect linkages.
Quantify uncertainty, confidence intervals, and perform model diagnostics (VIF, residual analysis, cross- validation).
Use multivariate and non- linear regression to examine interdependent behavioral relationships (e.g., writing acceleration vs. tone modulation).
Machine Learning & AI Integration
Experiment with speech emotion recognition, sequence models, or multimodal fusion networks that integrate handwriting + audio.
Use NLP and embedding models to analyze transcribed speech or open- ended answers for affective or cognitive insights.
Collaborate with engineers to prototype LLM- driven insight layers that summarize behavioral findings or explain patterns.
Develop predictive models to forecast engagement, confidence, or completion speed.
Research & Hypothesis Testing
Design and conduct experiments, quasi- experiments, or A/B tests to validate hypotheses and interventions.
Build reproducible research pipelines (Jupyter, MLflow, W&B) with version- controlled analysis and documentation.
Use statistical hypothesis testing (ANOVA, chi- square, t- tests, permutation testing) to verify observed trends.
Apply causal reasoning to determine which variables most strongly influence learning efficiency.
Data Infrastructure & Visualization
Document datasets, model assumptions, and findings in clear technical and narrative formats.
Optimize ETL workflows for handwriting and voice signals, ensuring high- quality data ingestion.
Create interactive dashboards or visualizations (Streamlit, Plotly, Tableau) to communicate insights intuitively.
Work with engineers to maintain clean, labeled, and reliable multimodal data pipelines.
Collaboration & Communication
Support leadership with metrics that guide pedagogy strategy, AI model improvements, and product direction.
Translate analytical outputs into actionable insights for curriculum and product design.
Present complex analyses in simple, visual, and narrative forms for non- technical stakeholders.
Partner with educators and product managers to interpret results in a learning context.
