Francis Paul C. Flores

• Engineered an automated Python and Plotly dashboard suite for key tracker studies, reducing client reporting turnaround by over 95% and saving analysts 20 hours weekly in manual reporting bottlenecks.
• Fine-tuned an XLM-RoBERTa-Large AI model to categorize hundreds of thousands of open-ended survey responses against established tracker codeframes; achieved an ~80% agreement rate with human coders while compressing turnaround time from weeks to hours.

• Architected reproducible R/Python analytical pipelines processing 5M+ rows of Hong Kong government data, automating EDA, feature engineering, model training, and reporting — standardizing the team's end-to-end research workflow and eliminating manual one-off scripts.
• Quantified population-level health risk factors across Hong Kong's 18 districts by applying supervised ML (GLM, Random Forests, SVM, Naive Bayes) and unsupervised segmentation (K-Means, Latent Class Analysis), directly informing resource allocation and intervention planning for the Food and Health Bureau.
• Modeled public health trends, intervention effects, and longitudinal risk trajectories using time series methods (ARIMA, VAR, GARCH, Prophet, LSTM) and panel models (mixed-effects, GEE), underpinning multiple publications across leading journals.
• Authored statistical methods and results sections for manuscripts in The Lancet, The Lancet Public Health, Nature Medicine, JAMA Network Open, and The Lancet Regional Health.
• Recognized with the HKU Knowledge Exchange Excellence Award (2024) for bridging academic research and real-world public health policy impact.
• Collaborated on study design, power calculations, and analysis plans; mentored junior researchers and students in statistical modeling and reproducible coding best practices.

• Championed the adoption of R across the analytics team (migrating from XLStat), establishing standardized RMarkdown workflows that boosted reporting reproducibility and reduced manual effort across the team.
• Implemented MaxDiff Scaling and Shapley Value Regression to quantify consumer preference drivers, delivering statistically rigorous, client-ready market research insights.
• Built predictive models, demand forecasts, and latent consumer segmentations to identify market demand patterns, directly informing clients' go-to-market and product positioning strategies.

Ni, M. Y., et al., & Flores, F. P., et al. (2020). Depression and post-traumatic stress during major social unrest in Hong Kong: a 10-year prospective cohort study. The Lancet, 395(10220), 273–284. doi:10.1016/S0140-6736(19)33160-5

Ni, M. Y., Canudas-Romo, V., Shi, J., Flores, F. P., et al. (2021). Understanding longevity in Hong Kong: a comparative study with long-living, high-income countries. The Lancet Public Health, 6(12), e919–e931. doi:10.1016/s2468-2667(21)00208-5

Shi, J., Leung, C. M. C., Chen, R., Xiao, X., Flores, F. P., et al. (2026). Interpersonal conflicts, social media use and depression associated with protests. Nature Medicine, 32(1), 224–230. doi:10.1038/s41591-025-04145-0

Lun, P., Ning, K., Wang, Y., Flores, F. P., et al. (2023). COVID-19 Vaccination Willingness and Reasons for Vaccine Refusal. JAMA Network Open, 6(10), e2337909. doi:10.1001/jamanetworkopen.2023.37909

Ni, Y., Flores, F. P., Lun, P., Ning, K., et al. (2024). Understanding paediatric COVID-19 vaccination during the pandemic: a prospective cohort and population-based registry study. The Lancet Regional Health – Western Pacific, 43, 100976. doi:10.1016/j.lanwpc.2023.100976