Post-doc: Probabilistic Disease Trajectory Modeling at Harvard/Harvard Medical School
Country/Region : Czech Republic
Website : https://seas.harvard.edu
Description
We're advertising an exciting postdoctoral opportunity to work with Harvard computer scientist and Harvard MD on disease discovery; this postdoctoral position is joint between Harvard SEAS and Harvard Medical School. We seek a recent PhD researcher with a strong background in probabilistic methods and timeseries modeling and an interest in healthcare applications.
-- The Project --
Discovering disease trajectories and subtypes is a critical step toward both understanding the underlying causes of a disease and guiding treatment. Our group has been developing methods to discover disease trajectories and subtypes from electronic health records and social media, using modeling approaches ranging from variants of hidden Markov models to dynamic topic models. However, there are still important challenges, ranging from making the models easier to interpret (in particular, by adding side+relational information) and algorithms for robust, scalable inference in settings with high dimensionality, complex missingness patterns, and limited computational resources.
The objective of this project is to develop and implement solutions to these challenges, with the goal being to be able to use such models for knowledge discovery for a variety of diseases. The successful completion of this project will require both algorithmic development and implementing the algorithms so that they can be applied to millions of patient records. There will be an opportunity to integrate the work into popular clinical research software (via assisting a production team). Data formats are all standardized and we have software for basic versions of our models, so this is a great opportunity to jump into a technical, impactful project without too much grunt work.
-- Requirements --
The applicant should have a strong machine learning background with a track record of top-tier research publications (e.g. JMLR, NIPS, ICML, AAAI, AISTATS, IJCAI, KDD). Expertise in probabilistic models and numerical programming in Python is a must, and experience with timeseries models and moderately sized data sets is a big plus.
-- Length --
The initial position will be for 18 months with the option of starting in January or February 2017, with the possibility of extending, depending on performance. Harvard is an equal opportunity employer.
-- To Apply --
Please send finale-AT-seas.harvard.edu and Paul_Avillach-AT-hms.harvard.edu your application consisting of
- subject line: [PDTM Postdoc Application]
- cover letter with research statement, names and emails for 3 references
- detailed CV
- 3 most relevant/top publications
-- Deadline --
Applications will be considered on a rolling basis through Dec. 1
-- The Project --
Discovering disease trajectories and subtypes is a critical step toward both understanding the underlying causes of a disease and guiding treatment. Our group has been developing methods to discover disease trajectories and subtypes from electronic health records and social media, using modeling approaches ranging from variants of hidden Markov models to dynamic topic models. However, there are still important challenges, ranging from making the models easier to interpret (in particular, by adding side+relational information) and algorithms for robust, scalable inference in settings with high dimensionality, complex missingness patterns, and limited computational resources.
The objective of this project is to develop and implement solutions to these challenges, with the goal being to be able to use such models for knowledge discovery for a variety of diseases. The successful completion of this project will require both algorithmic development and implementing the algorithms so that they can be applied to millions of patient records. There will be an opportunity to integrate the work into popular clinical research software (via assisting a production team). Data formats are all standardized and we have software for basic versions of our models, so this is a great opportunity to jump into a technical, impactful project without too much grunt work.
-- Requirements --
The applicant should have a strong machine learning background with a track record of top-tier research publications (e.g. JMLR, NIPS, ICML, AAAI, AISTATS, IJCAI, KDD). Expertise in probabilistic models and numerical programming in Python is a must, and experience with timeseries models and moderately sized data sets is a big plus.
-- Length --
The initial position will be for 18 months with the option of starting in January or February 2017, with the possibility of extending, depending on performance. Harvard is an equal opportunity employer.
-- To Apply --
Please send finale-AT-seas.harvard.edu and Paul_Avillach-AT-hms.harvard.edu your application consisting of
- subject line: [PDTM Postdoc Application]
- cover letter with research statement, names and emails for 3 references
- detailed CV
- 3 most relevant/top publications
-- Deadline --
Applications will be considered on a rolling basis through Dec. 1
Last modified: 2016-10-12 23:27:26