I view the Internet of Things (IoT) akin to the smartphone app market. How this market shapes out is presently up for grabs. Corporations are constantly trying to put stakes in the ground, but a great deal of the science and technology is missing for IoT to materialize into a successful industry. Apple enabled the ingenuity of app developers by building consistent infrastructure for developers to code and deploy on. Since then, we have a handful of other platforms, namely Android, and Windows 8. IoT apps may one day enjoy similar benefits, but they live in a much more complex space. The crossing between the physical and the cyber is as diverse as the world and as heterogeneous as manufacturers allow their imagination to carry them. This is where my research comes into play, the details of which will be articulated in my thesis.

I was part of a multidisciplinary team from Washington University Medical and Engineering schools, aided by an advisory committee of older adults. The purpose of this research is to develop and test a novel fall detection system for community dwelling older adults. Our research is novel in its approach and data sources. We are not trying to devise a fall detection algorithm per se, and we are not testing our ideas on students. Instead, we characterize older adult gait and movement patterns using machine learning, enabled by volunteers from the target population. Our plan is to build a fall detection system that is non-intrusive, omnipresent, and has high efficacy.

This research is in its early stages and is looking for additional funding.

Wireless sensor networks can play an important role in improving patient care by collecting continuous vital signs and providing clinical decision support. Such a system has the potential to save lives with real-time alerts sent in-between nurse's bedside visits or physician medical rounds (in non-critical care settings, these in-between periods may be as long as 8 hours). Telemetry can continue during treatment in other hospital departments or can continue at home.

We, a multidisciplinary team of Physicians, Artificial Intelligence (AI), and System researchers developed a two-tier system, which enables early diagnosis of impending clinical deterioration. In the system's first tier, machine learning algorithms data mine clinical data of past patients. The second tier is focused on quality delivery of vital signs in a hospital environment. The Early Warning System (EWS) we developed identified patients at risk of clinical deterioration. Its alerts were highly associated with ICU transfers, and the algorithms alerted personnel hours prior to the onset of conditions.

These days research furthers to home monitoring. This research direction may allow early discharge from hospitals, augmented with at-home monitoring, reduced readmission rate and improved recuperation. I am not involved with this part of the project.

Human beings, most life forms, and cameras use only the visible and unpolarized part of the electromagnetic spectrum to see or construct images. The focus of this research rotation was to empirically verify whether polarized data can be used to form better images under difficult light conditions.

The picture on the left shows an airplane that would otherwise be completely veiled by the dark night. Click the image to see the light conditions under which this image was reconstructed.

Developing a mathematical model of a system is a key stepping stone to building a high performance or parallel system. But often, modeling is absent from the design process due to the complexity of the model development. As the system evolves, even when models were developed, they may diverge from the true behavior of the system because their complexity makes them hard to update.

We researched novel approaches that greatly simplify the development of such mathematical models and the computations involved, rendering our techniques palatable for designers.

To test our theory, we modeled a hybrid system called Mercury BLAST, in which parts of NCBI BLAST are implemented on FPGA. This high performance version of BLAST runs 10-30x faster than NCBI BLAST on a single AMD Opteron CPU core, while delivering 99+% of the same results.

Geographically distributed organizations that wish to enable varying degrees of sovereignty to different parts of the organization are faced with data coherency and synchronization challenges. I developed a novel data transport system that enhanced organization agility, allowed local decision making using timely and accurate data, and provided decision-makers with clear visibility of data aggregated from the city, regional, national, and global headquarters.

Visual Office (VO) is an ERP system that we developed on top of this distributed data objects underpinning. VO solved the problem of distributed heterogeneous data stores, some of which were data-bases and a diverse set of other data sources.

Post research, VO was implemented in large organizations. It enjoyed particular success with distributed businesses, such as franchise management, chain stores, and distribution of goods. VO offered modules such as inventory control, order management, invoicing, accounts receivable, contact information, and business forms.