CORPORATE KEYNOTE SERIES
Christopher Geiger is the engineering director and chief engineer of the Lockheed Martin Enterprise Sustainment Solutions (ESS) market segment. His research interests include aerospace and defense sustainment information systems, automated test systems, and the resilience and cyber security of ad hoc networks. He received a master's in business administration and bachelor’s degrees in electrical engineering and chemistry from the University of Florida. He is a licensed Professional Engineering (PE) in Florida and
Texas and a UK Chartered Engineer (CEng). Mr. Geiger is also on the board of directors of MidFlorida Credit Union (in the top 100 in the US), a local elected official, and chair of the Florida Professional Engineers in Industry. In addition, he is a senior member of the American Institute of Aeronautics and Astronautics (AIAA) and IEEE as well as a member of the Royal Aeronautical Society (RAeS). Mr. Geiger resides in Orlando, FL with his wife Lisa and three children where you can often find him canoeing and hiking in the wilds of Central Florida.
Title of the Talk: What do Aerospace and Defense Sustainment Information Systems have to do with the Latest Trends in Technology?
Abstract: There is an ebb and flow to the transfer of technology between defense and commercial industries. Many current hot topics such as machine learning, cloud computing, mobile devices, and autonomous vehicles receive significantly more commercial R&D funding than from defense. However, aerospace and defense sustainment information systems have had a head start over commercial Internet of Things (IoT) and the related topics above. This talk outlines surprising commonalities, shared lessons, and some potential futures from aerospace and defense sustainment as we all hurtle toward an ever more interconnected world.
Chiman Kwan received his BS (honors) with major in electronics and minor in mathematics from the Chinese University of Hong Kong in 1988, and MS and Ph.D. degrees in electrical engineering from the University of Texas at Arlington in 1989 and 1993, respectively. He is the founder and Chief Technology Officer of Signal Processing, Inc. and Applied Research LLC, leading research and development effort in real-time control, chemical agent detection, biometrics, speech processing, image fusion, remote sensing, mission planning for UAVs, and fault diagnostics and prognostics.
From April 1991 to February 1994, he worked in the Beam Instrumentation Department of the SSC (Superconducting Super Collider National Laboratory) in Dallas, Texas, where he was heavily involved in the modeling, simulation and design of modern digital controllers and signal processing algorithms for the beam control and synchronization system. He later joined the Automation and Robotics Research Institute in Fort Worth, where he applied intelligent control methods such as neural networks and fuzzy logic to the control of power systems, robots, and motors. Between July 1995 and April 2006, he was the Vice President of Intelligent Automation, Inc. in Rockville, Maryland.
Over the past 27 years, Dr. Kwan has served as Principal Investigator/Program Manager for more than 115 competitively selected projects with total funding more than 36 million dollars from various government agencies and private companies such as Ford, Motorola, Boeing, Honeywell, and Stanford Telecom. He has 15 issued and pending patents, 50+ invention disclosures, 310+ journal and conference papers, and 450+ proprietary technical reports. He received numerous awards and recognitions from NASA, US Navy, US
Air Force, and IEEE.
Title of the Talk: Enhancing Safety of UAVs in National Airspace
Abstract: Unmanned Air Vehicles (UAV), also known as Unmanned Air Systems (UAS), are gaining more attention
in recent years. Some potential commercial applications may include cargo transfer between major cities, package and food delivery to individual households, etc. However, it is well-known that UAVs are much less reliable and have far more accidents than manned aircraft. This is probably one of the most important reasons that FAA is hesitant to open up the national airspace (NAS) and imposes tight restrictions to UAVs. Reliability of UAVs can be strengthened using durable engines and communication equipment, strong structural materials, advanced conditioned based maintenance and structural health monitoring procedures, accurate fault diagnostic algorithms, and robust and fault tolerant controllers. Despite the above measures, some equipment failures such as engine and communication equipment failures may still occur. In this talk, we present some recent research results done by our team to deal with major contingencies such as lost communications and engines failures. In particular, contingency planning software prototypes have been developed that can deal with those aforementioned major contingencies. Architectures of contingency planning software and some exemplar application scenarios will be discussed throughout the talk.
Erwin Bellers received his M.Sc degree in Electrical Engineering with distinction from University of Twente in the Netherlands in 1993 and his PhD degree in Electrical Engineering from Delft University of Technology in 2000. Erwin joined Philips Research in Eindhoven, the Netherlands, in 1993 as a Researcher and mainly conducted research in video enhancement, de-interlacing and frame-rate conversion. In 2000 he joined Philips Research USA in New York as a senior Scientist where he led a research-team focusing on spatial resolution enhancement.
In 2002 Erwin moved to Silicon Valley to join Philips Semiconductors as a Senior Video Architect, and continuing into NXP Semiconductors in 2006. Erwin further innovated on the algorithm and architecture of frame-rate conversion which became a successful product range of NXP Semiconductors.
With the merger of NXPs TV business and Trident, Erwin became responsible for video algorithm innovations within Trident. In 2010 the TV business of Trident was acquired by Sigma Designs and Erwin continued in several roles including director, senior principle and Fellow. In June 2018 the TV business of Sigma Designs was acquired by the startup V-Silicon. Currently, Erwin is responsible for the video innovations and leading the Video Innovation Team in his role as Fellow Engineer.
Erwin was invited as a co-author for the Proceedings of the IEEE, and he won the second price in the ICCE Outstanding Paper Awards program in 1997 and 2006, and the first price in 2009. Erwin has published over 50 papers, 1 book, 1 book chapter, and holds over 50 patents and patent applications.
Title of the Talk: Challenging Algorithm and Architecture design in the realm of Consumer Electronics.
Abstract: The Consumer Electronics (CE) market is very challenging. Bounded by tremendous price erosions, short
time-to-market cycles, high level of Silicon integration, high customer expectations, constantly changing requirements, expected level of adaptation over its lifetime, level of innovation, etc., creates an interesting landscape which stimulates creativity and innovation at all levels.
In television sets the video resolution has seen an increase from 1K to 2K to 4K and now pushing towards 8K pixels, from Standard Definition Range to High Dynamic Range (HDR), from a relative small to a large color space, from single to quad CPU cores, from simple graphics to more complex graphics. In the presentation we will address various challenges from a television System-on-Silicon (SoC) providers point of view, discuss some key video technologies from algorithmic and architectural perspective, and some opportunities and challenges ahead.
Yuhan Cai is an Engineering Manager at LinkedIn, leading a team of scientists/engineers to build machine learning infrastructure products, facilitating the robust, efficient and straightforward application of machine learned capabilities. Before LinkedIn, he has worked at bigger companies like Apple, Microsoft and Amazon, as well as smaller startups, focusing on scalable big data platforms and AI advancement. He received a Bachelor and Master of Computer Science from the University of British Columbia in Canada, and he was the winner of ACM SIGMOD Best Paper and Best Demo Awards.
Title of the Talk: ReMix: Simplifying Workflows in Machine Learning Infrastructure
Abstract: Our mission at Machine Inference Infrastructure is to build easy, scalable and operable machine learning to connect the world’s professionals to the resources they need, and our vision is to leverage Artificial Intelligence at massive scale to bring real economic opportunity to the world. In this talk, I’m going to present one of our machine learning infrastructure products, ReMix, which is a generic workflow orchestration solution for building mid-tier services. ReMix was originally conceived as the next generation federation solution unifying LinkedIn federation infrastructure. The lessons learned from the federation use cases applied generally to many Relevance verticals, thus ReMix has been built as a more generalized solution for building Relevance mid-tier services. It provides a declarative Java API for composing workflows from operators and from other workflows. It includes convenient support for making asynchronous calls to external REST resources, batching/caching data fetches needed by multiple operators in a workflow, and other performance and robustness enhancing workflow optimizations. It has reached General Availability in June 2018 and been running in production with Ranker Relevance (recommendation) Models/Endpoints and search blending since 2018.
Joel Jones received his PhD from the University of Illinois in 2002 for work on machine-independent ahead-of-time Java optimization. He has worked both in academia and in industry for such organizations as the University of Alabama, Mills College, Apple Computer and for startups. He has worked for Cavium (now Marvell) for the last five years, and has contributed to GCC, LLVM, and has helped meet other "challenges".
Title of the Talk: Software Challenges in Introducing a New Architecture
Abstract: There has been only one predominant architecture in both cloud and HPC computing for decades. To challenge this mono-culture, competitive hardware was just the beginning. A host of software challenges have been addressed by many actors working in concert. I present my perspective as an individual contributor and leader of a tightly focused toolchain team on the challenges that we have dealt with in concert with various partners. Our areas of focus have included compilers, operating systems, performance measurement and improvement, and the broader eco-system (and standards for machine screws). I will also discuss ongoing efforts in these and related areas.
Chuyen Luong received an Engineering degree in School of Electronics and Telecommunications from Hanoi University of Sciences and Technology, Vietnam in 2012; She got MS degree in Degree in School of Electronics and Computer Engineering, Chonnam National University, South Korea in 2014. She is currently working as a software engineering in Viettel Network Technologies Center (VTTEK), Viettel Group, Vietnam. Her research interests are mainly in in-memory database system, distributed database system, NoSQL.
Title of the Talk: A Concurrency Control Method in Distributed Multi-core Database Management System
Abstract : A concurrency control method is proposed to guarantee the serializable isolation level based on ordering logic commit timestamps when processing many transactions simultaneously in distributed multi-core database management system. This method calculates a logic commit timestamp assigning to a transaction processing based on timestamp ranges of data items in operations of transaction in server cores. A distributed transaction processing consists three phases including read phase, prepare phase and commit phase as follows:
- - Read phase: Transaction read data from servers.
- - Prepare phase: Validate transaction based on calculating a suitable logical timestamp range of data items in cores of replica servers. Moreover, if the transaction validation is successful, then returns calculated timestamp range to client.
- - Commit phase: If at least an operation is prepared unsuccessfully or client cannot compute a suitable timestamp range from operators’ timestamp ranges, client will abort that transaction to servers. Otherwise, if client can find out a suitable timestamp range from operators’ timestamp range, client will commit transaction to servers with a commit timestamp in that found range.
This method is used the logical timestamp without physical timestamp which need to be synchronized between servers. Furthermore, it can avoid the bottleneck problem in Sundial protocol, deadlock problem in 2PL technique, or reduce aborted transaction in MaaT.