Long time fan, first time poster - thought I’d do an intro for me and my team.
I’m Jon, a data scientist at UKHSA. I currently lead our short-term forecasting and nowcasting (affectionately termed “pancasting”) workstream within the infectious disease modelling team led by Tom Ward. There are about 4 of us working on these nowcasting/forecasting problems.
Our primary work is delivering short-term forecasts for pathogens that cause healthcare pressures over winter. This year we are producing ensembles for COVID-19, influenza & RSV admissions + bed occupancy, as well as norovirus cases. Overview here.
We deploy and maintain our own ensemble that goes out to users, all done in-house which means I’m very interested in how to design optimal ensembles. The team works in R and Stan, we do a lot of work on GAMs (mgcv). We are a very stats focused team, so I am keen this year to explore incorporating more mechanistic components into our ensembles. As we are very applied side a lot of our focus is on approximate methods and evaluation of different approaches in applied contexts.
In addition, we provide modelling support during outbreaks (in 2024 we explored pertussis and delivered a measles nowcast).
We do not do as much open-source software development, or pure methods research as I would like - but I am very keen to support those doing so. If having an opinion from someone within the system would be helpful, please do get in touch. I am always happy to have a call.
About me more personally, I am currently working on a PhD by publication and couch to 5k (one is going better than the other). I’ve worked in government for ~5 years at UKHSA and ONS.