Mike Rosen Email & Phone Number

Edmonds, Wa

I am the lead engineer for CalMAN, the gold standard for Video Monitor Calibration software. It's claim to glory is it's ability to automate the calibration of video monitors. It's quite complex and makes great use of WPF to provide a highly interactive UI while talking to many displays and instruments at the same time. This application is the mainstay of the company and my focus is on quality, stability and maintainability. When I arrived in 2018, this key product had been un-maintained for over two years. Since then, it's not only seen multiple, high-quality releases, it has a regularly-executing unit test suite that demonstrates a readiness-to-ship that removes much of the uncertainty in release cycles.

Greater Seattle Area

Azimetry is a small Lidar Production company. The main line of business is the largely manual process of Lidar classification / 3d feature extraction. The company seeks to automate some of this and I’m the principal engineer researching and implementing an ML-based classifier. My time is split between reviewing the most recent academic literature in this area and implementing and evaluating that research. We’re working in Linux, mostly with Python (SKLearn, Numpy, LASPy, TensorFlow) but also C++ (PDAL, GDAL, Displaz). Other responsibilities include proving associated workflows in QGIS as well as higher level collaboration via industry, open source and academic forums.

Seattle

I provided technical guidance on foundation code and integration points and was responsible for key enhancements and refactorings. Externally, I represented the company at several industry forums including the ASPRS (I’m the VP of our local chapter), OGC and OSGeo, Here are a couple of highlights:- In 2016, our Geospatial Search product required enhancement to be able to index DIMAP (metadata associated with SPOT satellite imagery). I enabled this via our existing GDAL-based architecture. However the customer noted that while some DIMAP files now worked correctly, others did not. I diagnosed the failure; it was due to the way that the offending images had their geographic coordinates specified. To solve the problem, I extended GDAL’s DIMAP support to include RPC-based geo-referencing. I submitted a patch back to the community. - By 2009, our principal product had evolved into a Windows Forms / Managed C++ desktop application. At 10 years old its mixture of business logic and UI logic has made it brittle and hard to maintain. Far and away the most important product the company ships, re-writing it was not an option. To significantly enhance the UI as well as make a path forward for future maintainability, I replaced about a third of the UI with WPF controls. This involved re-factoring portions of the business logic into interfaces which could be defined on the WPF side ("dependency inversion"). It shipped this way for another five years. In 2014, we were able to further the refactoring, by including a proper data model against which we could use WPF data bindings.- LAS is the primary format used to exchange LiDAR data. The ASPRS LAS subcomittee (of which I am member) is responsible for maintaining the associated specification. I articulated the significant deficiencies associated with the legacy geo-referencing mechanism and ensured it was addressed in the subsequent specification revision.

I led the team that delivered the industry's first Lidar-compression application. As the project manager, I mitigated risks like selection of compression algorithms (researched as part of development cycle!), feature-vs-schedule trade-offs and buy-vs-build decisions (resulting in sponsorship of the open source liblas project). As one of two developers, I built the WPF application that managed compression/decompression jobs and visualized lidar data. Here's a link to an engineering overview of that project (including screenshots):(http://www.lizardtech.com/blog/2009/lizardtech-engineering%E2%80%99s-introduction-to-lidar-compressor/)I added measurement capability to the company's primary viewing application, a web-browser plugin. The development environment here was C++, COM, the Netscape plugin API and GDAL (open source geo-spatial library).

I ran a development team of between 3 and 6 engineers who were responsible for a scanned-image product line called "DjVu" based on research technology acquired from AT&T Labs. In an era before AJAX, we were able to stream 300-page manuals over low-bandwidth connections and give a "just like paper" user experience. This was adopted by Samsung, Kubota and Hitachi. Our DjVu Browser Plugin which enabled this experience got > 1 Million downloads. Google, "samsung manuals djvu" to try it out. I made significant technical contributions to the codebase (x-platform C++) as well as managing three major release cycles.We won the "Complete New Yorker" project: In 2005 our technology turned 70 years of old magazines into #23 on Amazon. The Complete New Yorker is exactly what you think it is: every page of every issue ever published. It’s all DjVu under the covers. I was the principal technical contact with Condé Nast's vendor, optimizing encoding parameters and troubleshooting SDK issues. A few years after that, the same vendor came back to us, re-licensed the technology and did the same thing with "Playboy Cover to Cover" and "Rolling Stone Cover to Cover." All three are still available on Amazon.A surprisingly large part of maintenance here was dealing with irregularities in specific raster files. Formats like TIFF, PNG, JPEG and PDF all have many inconsistently implemented edge cases and needed addressing using associated open source libraries (libTiff, libPNG, LibJPEG and GhostScript here). Examples included things like MultiPage tiff files, the "PhotometricInterpretation tag" (does 1 mean black or white) and many, many PDF encoding irregularies.

Over a 3 month period, I lead a team of Systems Analysts who delivered the high-level software architecture for a wafer fab startup in Singapore. The startup was a joint venture between established foundries who had different systems for running a wafer fab. Our deliverable was a solution architecture that identified what systems would allow proper operation of the shared fab and what integration pieces were missing. It covered everything from the capacity of the network cables to the cost allocation models used by the ERP systems.For 18 months, I worked for EDS’s Metrics and Estimating Centre estimating the cost of prospective projects by doing a statistical analysis of a database of past projects. We used Excel and SPSS to do the analysis.For 24 months, I automated the machines that turn Silicon wafers into memory chips at Fujitsu’s Gresham Manufacturing Division. This involved writing “device drivers” that integrated the “tools” into the fab’s workflow control system and working directly with the tool’s manufacturers to resolve standardization issues. The programming part was a C-like language that ran under Solaris. The analysis required engineer-to-engineer meetings with the associated manufacturers.

I did requirement analysis and programming on a US Navy project management system. The front end was an early cross-platform GUI builder called JAM. The back end database was Oracle 7.

Master’S Degree, Software Engineering

Bachelor’S Degree, Mathematics