Eugene Lavely writes,

I work at BAE Systems AIT here in the Boston area. We have a very interesting research program for imaging building interiors that is just starting. I believe the unique features of this inverse problem leads one to MCMC methods such as RJMCMC. The specific statistical techniques that we would like to apply are described in the second paragraph of the ad below. We seek candidates with outstanding expertise in these areas. Our position is available for immediate start. BAE has a very academic atmosphere,and so academic scientists fit in here very well, and have significant opportunities to develop and apply innovative research.

Trans-dimensional Inverse Problems and Markov Chain Monte Carlo

We [BAE Systems AIT] are seeking candidates with interest in global optimization and posterior simulation for complex, physics-based inverse problems. This position opening is an exciting opportunity to be on the ground floor of a new program start to image 3D building interiors with microwave sensor technology. This difficult problem requires coupling of fast (but accurate) electromagnetic forward modeling tools that capture ray, diffractive and diffusive wave behavior with an inference scheme that addresses the numerous problem challenges. You will work closely with domain experts in high-performance computing and electromagnetic modeling to develop the inferencing and simulation algorithms described below. A key objective is the design and implementation of a closed-loop model estimation scheme for estimation of the layout of building interiors using data from a collection of mobile, networked, microwave sensors that may operate in monostatic and bistatic modes. Closed-loop refers to the desired ability of cooperative sensor behavior and adaptivity so that data collection plans may be adjusted to ensure data collection supports optimal inference. Optimal experiment design and/or multi-objective function optimization may play a role here. You will participate in spiral cycles of algorithm design and development for testing and parallel implementation on a Beowulf cluster. As the program continues, and increasing amounts and sources of data are acquired, you will use these inputs to refine and improve the integrated processing approach. You will also assist in the design of data collection experiments to provide data suitable for proof of concept and system demonstration capability.

The inference problem is a trans-dimensional inverse problem in which the model space and associated parameter space must be simultaneously explored since the model-order (building CAD model) is unknown. There are methods that have been developed in the Bayesian inference community such as Reversible Jump Markov Chain Monte Carlo (RJMCMC) and trans-dimensional simulated annealing (TDSA) that, in principle, can be adapted to this tomographic problem class. In applying these methods it will be important to introduce techniques that expedite exploration of the model space, accelerate chain convergence and suppress random walk behavior. Possibilities include parallel and simulated tempering, dynamic weighting, population-based Monte Carlo and Metropolis-coupled MCMC. Experience in the design of efficient proposal distributions and model move processes appropriate to the physical requirements of the problem, and that satisfy conditions of reversibility and detailed balance is a plus. Polygonal random fields (e.g., as governed by the Arak process) are also likely to play a role in our approach so understanding of the associated probability measures is also a plus. It will be of interest to pose the inverse problem as one of hierarchical Bayesian estimation, so experience with graphical models and directed acyclic graphs is also helpful. Ideally, the candidate also has modeling and inverse theory experience in one or more imaging modalities such as electromagnetic, seismic, or acoustic.

U.S. Citizenship is desired, but highly-qualified candidates with Green Card status are also invited to apply. BAE Systems AIT is located in Burlington, MA just 10 miles outside of Boston, and near the historical towns of Lexington and Concord, MA. Additional company information can be found at http://www.ait.na.baesystems.com/primary/index.htm. Review of applications will begin immediately, and continue until the position is filled. Interested candidates should contact Eugene Lavely via email at [email protected] or by phone at 781-273-3388 ext. 294.

This looks interesting to me. Maybe a job I would’ve been interested in had it existed when I got my Ph.D.