Curriculum Vitae
Alexander Feldman

October 24, 2022

Date of Birth:

September 17, 1977

Tel.:

+1 650 6444111

email:

alex@llama.gs

URL:

http://llama.gs/personae/alex/

Bio

Alexander Feldman is a research scientist with significant contribution to artificial intelligence, quantum computing, computer science, and engineering. His current work is on algorithmic methods for combinatorial optimization and their use in diagnostics, design, and manufacturing. Dr. Feldman’s work in AI has primarily been in model-based diagnostics, applications of satisfiability, quantum computing and machine learning with contributions to the automation of electronic design (digital, analog, and mixed-mode), robotics and localization, constraint optimization, and program synthesis. He has published more than sixty articles in leading peer-reviewed journals and conferences. Until recently, Alexander Feldman was a member of research staff at the Palo Alto Research Center (Xerox PARC). Prior to joining PARC, he was a Ph.D. student at Delft University of Technology. The topic of his Ph.D. thesis was “Approximation Algorithms for Model-Based Diagnosis”. Dr. Feldman has ten accepted patents and a few more are pending acceptance.

Education
9/2005 – 5/2010

Ph.D. (cum laude), Computer Science
Delft University of Technology, The Netherlands

Thesis: Approximation Algorithms for Model-Based Diagnosis

Advisor: Prof. Arjan van Gemund

9/2002 – 9/2004

M.Sc. (cum laude), Computer Science (Technical Informatics)
Delft University of Technology, The Netherlands

Thesis: Hierarchical Approach to Fault Diagnosis

Advisor: Prof. Arjan van Gemund

9/1997 – 6/2000

B.Sc., Computer Science
UE Varna, Bulgaria

Employment
9/2014 – 9/2022

Member of Research Staff
System Sciences Laboratory, Model-Based Reasoning Area
Palo Alto Research Center, Inc. (Xerox PARC)
California, USA

6/2013 – 9/2014

Founder and President
General Diagnostics, Delft, The Netherlands

6/2013 – 9/2014

Technical Consultant
Nspyre, Eindhoven, The Netherlands

6/2012 – 6/2013

Postdoctoral Research Fellow
Complex Systems Laboratory
University College Cork, Ireland

6/2010 – 6/2012

Visiting Postdoc
Radio Frequency Integrated Circuit Group
Ecole Polytechnique Fédérale de Lausanne (EPFL)

6/2010 – 6/2012

Visiting Postdoc
Distributed Intelligent Systems and Algorithms Laboratory (DISAL)
Ecole Polytechnique Fédérale de Lausanne (EPFL)

6/2010 – 6/2012

Postdoc
Institute of Information & Communication Technology
Haute Ecole d’Ingénierie et de Gestion du Canton de Vaud, Switzerland

5/2008 – 9/2008

Visiting Researcher
Intelligent Systems Laboratory, Embedded Reasoning Area
Palo Alto Research Center (PARC), Inc.
California, USA

9/2005 – 5/2010

Doctoral Research Fellow
Embedded Software Laboratory, Department of Software Technology
Faculty of Electrical Engineering, Mathematics and Computer Science
Delft University of Technology, The Netherlands

4/2005 – 9/2005

Software Architect
Science and Technology BV, Delft, The Netherlands

9/2001 – 4/2005

Senior Programmer
Market Risk Management, ING Bank, Amsterdam, The Netherlands

7/2000 – 9/2001

Senior Programmer
Zend Technologies Ltd., Ramat Gan, Israel

Honors and Awards

Ph.D. cum laude

M.Sc. cum laude

Best paper award at the First International Conference on Prognostics and Health Management 2008 (PHM’08)

Gold Leaf certificate at the Seventh Conference on Ph.D. Research in Microelectronics & Electronics 2011 (PRIME’11)

Best paper award at the Eleventh Annual Conference of the Prognostics and Health Management Society (PHM’19)

Project Involvement
Digital Deisgner

Boolean circuits are fundamental in computer science and microelectronics. They can be used both for the synthesis of digital Integrated Circuits (ICs) such as micro-processors and for the modeling and analysis of software. One can also use Boolean circuits for machine learning. Although there exist methods and algorithms from model-checking, satisfiability, Satisfiability Modulo Theory (SMT) and related AI sub-disciplines, designing and analyzing digital circuits has been challenging.

In this fundamental research project, I have designed the first class of algorithms for multi-output Boolean circuit synthesis that are sound, complete, optimal, and, most importantly, general: they can synthesize a large class of digital circuits such as micro-processors, or quantum circuits, or sorting networks. The results led to a major journal publication, several invited talks (one at Stanford and another one at a major AI conference), invention submissions, and the ideas entered the portfolio of core competencies of my lab.

In the mid-future, the algorithms developed as part of this project should help VHDL and Verilog programmers, IC designers and software engineers to create and optimize part of their code automatically.

PARC FPAA

In this fundamental project I have designed and implemented a SPICE model of a Field-Programmable Analog Array (FPAA). FPAAs are the analog equivalents of Field-Porgammable Gate Arrays (FPGAs). This project led to a paper in a major conference and a response of a DARPA RFI.

The FPAA model that I am working on can be used for machine-learning, where the basic elements are not neurons but computational analog blocks with op-amps and configurable op-amp feedback.

The FPAA should give us better methods for simulation of analog electronics and for machine learning.

Fluidic Test-Bed

In this project, I have designed and implemented a fluidic system. By using the physical test-bed I have collected data for various fault-modes of the fluidic test-bed (for example a leaking proportional valve). I have analyzed the collected data with a number of model-based and machine-learning-based methods.

CBMx

We have designed and implemented a Condition Based Maintenance (CBM) platform for PARC. The purpose of this platform is to provide component-based rapid prototyping environment for building custom prognostic, diagnostic, and sensor-placement solutions. The platform provides automated analytics for optimal diagnostic and prognostic decision making, comparison of algorithms, diagnostic metrics, visualization and Supervisory Control and Data Acquisition (SCADA) interfacing, code generation for diagnostics in control, instrumentation and data cleansing, signal processing, and others. The CBM platform can service electrical, thermal, mechanical, or hybrid systems. An important application of the PARC CBM platform is to diagnose thermodynamic systems. My responsibility is to prepare a CBM use-case for thermodynamic cyber-physical system, to work on the diagnostic design aspects of the framework, to design novel algorithms and metrics and to improve state-of-the-art in diagnostic and prognostics within the framework.

CATO

CATO is a Dutch national programme (one of the participants is the Faculty of Civil Engineering and Geosciences at Delft University of Technology) whose aim is to study mechanisms for underground CO2 capture, transport and storage. My role was to support one of the experiments and to develop VHDL/LabVIEW instrument for preprocessing and storing of large amount of measurement data. The instrument allows sampling of up to 32 analogue signals with soft-adjustable sampling rate and acts as an averaging oscilloscope to improve the signal-to-noise ratio of the input signal. See http://www.co2-cato.org/ for more information.

Lydia-NG

Lydia-NG is a framework for Model-Based Diagnosis and is a continuation of my doctoral research. Lydia-NG bears resemblance to products such as Dymola and Rodon, however, it targets the automation of the design and implementation of decision support and diagnostic systems. The task of diagnosing a system is typically more difficult than simulation as it requires multiple simulations for various parametric values and advanced analysis for mode identification. Lydia-NG provides modeling and scenario languages (and compilers); many state-of-the-art libraries for simulation, diagnosis, and active testing; tools and component libraries. Lydia-NG provides own optimized simulation engines similar to the ones in the Spice circuit simulator or in Modelica. See http://general-diagnostics.com/ for more information.

UWB Localization

I have participated in this large National Centers of Competence in Research (NCCR) project during a two-year postdoc in Switzerland. The idea of the UWB project was to capitalize on the research and know-how on Ultra Wide Band Technology as well as multi-robot distributed search and localization techniques acquired in previous MICS phases. One of the main goals was to build a system that allows a team of mobile robots to locate themselves and other robots with high precision (order of a cm) very frequently (maybe once per second) and securely, in order to perform collaborative, such as distributed search, coverage, or mapping. The project was focused on distributed algorithms that can be efficiently implemented and on development of low power implementations on integrated circuits. I was responsible for the design and implementation of the receiver firmware and data acquisition algorithms. This project has been carried-out in collaboration with researchers from Ecole Polytechnique Fédérale de Lausanne (EPFL).

Genius

Decision support system for diagnosis satellite electrical power systems. Genius takes model-based diagnosis one step closer to the end-user by analyzing the real-world case of the Goce satellite. I have applied model-based diagnosis and active testing to data simulated with the Simsat ESA operational simulator. The results showed very good diagnostic performance (measured by performance metrics) which convinced end-users that model-based diagnosis and active testing is a mature technology, ready to be used in a wide-class of real-world systems. This project has been awarded by the European Space Agency under the Innovative Triangle Initiative program.

DXF

The DXC Framework (DXF), developed jointly with NASA Ames and PARC, is a collection of programs and APIs for running and evaluating diagnostic algorithms. DXF allows systematic comparison and evaluation of diagnostic algorithms under identical experimental conditions. The key components of this framework include representation languages for the physical system description, sensor data and diagnosis results, a runtime architecture for executing diagnostic algorithms and diagnostic scenarios, and an evaluation component that computes performance metrics based on the results from diagnostic algorithm execution.

Lydia

Lydia implements a big number of algorithms developed through my doctoral work. Lydia stands for Language for sYstem DIAgnosis and it is a modeling language and a reasoning tool-kit biased (e.g., there is support for health modeling) towards model-based fault diagnosis. One of the objectives of Lydia is to to implement novel algorithms which will push the frontiers of model-based diagnosis allowing efficient reasoning over larger systems. Responsible for the framework and modeling language design and implementation and the development of fast algorithms for model-based diagnosis.

Finesse

The project Finesse (Fault dIagNosis for Embedded SyStems dEpendability) aims at the improvement of the accuracy of fault diagnosis when applied to electromechanical systems such as the Paper Handling Systems of Océ Copiers. The challenges in fault diagnosis are to infer maximum diagnostic information on the operational status of software and hardware components from a typically limited amount of (noisy) observations. Responsible for the modeling of the system and the design of algorithms for active testing, recovery and prognosis.

DIF

The Diagnosis Interchange Format (DIF) is an XML-based interchange format for Model-Based Diagnosis (MBD). Its main purposes are to allow sharing of diagnostic models, observation data and fault hypotheses, and to facilitate empirical comparative study of the performance of existing and future MBD implementations. Responsible for the DIF schema design and the construction of MBD benchmark suite.

Professional Activities
PC Member

Thirteen AAAI Conference on Artificial Intelligence (AAAI’20)
Thirteen AAAI Conference on Artificial Intelligence (AAAI’16)
International Conference on Prognostics and Health Management 2015 (PHM’15)
International Workshop on Principles of Diagnosis 2015 (DX’15)
International Workshop on Principles of Diagnosis 2014 (DX’14)
International Workshop on Principles of Diagnosis 2013 (DX’13)
European Conference on Artificial Intelligence 2012 (ECAI’12)
International Conference on Principles of Knowledge Representation and Reasoning 2012 (KR’12)
International Workshop on Principles of Diagnosis 2011 (DX’11)

Reviewer

Journal of Vibration and Control
Journal on Artificial Intelligence (AIJ)
Journal of Universal Computer Science (JUCS)
Journal on Systems, Man and Cybernetics (SMC)
IEEE Transactions on Reliability (TREL)
International Workshop on Principles of Diagnosis 2010 (DX’10)
International Workshop on Principles of Diagnosis 2009 (DX’09)
International Joint Conference on Artifical Intelligece (IJCAI’13)
International Conference on Prognostics and Health Management 2011 (PHM’11)
International Conference on Prognostics and Health Management 2008 (PHM’08)

Organizer

International Workshop on Principles of Diagnosis 2013 (DX’20)
International Workshop on Principles of Diagnosis 2013 (DX’13)
Workshop on Diagnostic Reasoning and Model Analysis at European Conference on Artificial Intelligence 2012 (ECAI’12)
Third International Diagnostic Competition (DXC’11)
Second International Diagnostic Competition (DXC’10)
First International Diagnostic Competition (DXC’09)

Teaching Experience
7/2011 – 7/2011

Artificial Intelligence, California State University, Long Beach
Part of the summer university program which is an initiative launched in 2006 by the Board of Higher Education of the Canton of Vaud together with several partner universities.

9/2007 – 10/2007

Model-Based Computing, Delft University of Technology
Teaching assistant, but also designed the course and gave most of the lectures. This was an optional first-year M.Sc. course and was attended by approximately forty students.

2/2007 – 3/2007

Model-Checking, Delft University of Technology
Teaching assistant

Technical Skills

Proficient

Linux, Solaris, IRIX, HP-UX, Windows, ChibiOS/RT

C/C++, Boost, Qt, Python, PHP, Flex/Bison, LaTeX

MPI, PVM, CUDA, various other parallel programming frameworks

scikit-learn, Tensor-Flow

Sybase, Oracle, MySQL, PostgreSQL, various other databases

Maple, Matlab, Modelica (Dymola and Open Modelica), LabView

Digital Signal Processing, Kalman filters (also Extended and Unscented)

Lumped-Parameter physics modeling with Modelica

Control methods

Familiar

Java, Perl, Tcl/Tk, Prolog, Lisp, Pascal, Fortran

Familiarity with a number of various other ML frameworks

VHDL, Verilog

Hadoop

Various software engineering technologies such as version control, UML, SysML, etc.

Electronics design, analog to digital conversion, instrumentation, experimental design

Citizenships and Work Permits

Dutch citizen

US permanent resident

Languages

English, Bulgarian, Russian (intermediate), Hebrew (basic), Dutch (intermediate)

References

References will be available upon request.