Algorithm Group Manager at Rail Vision Ltd

Manager of  the Algorithm R&D Group (5 teams, overall), responsible for SotA solutions to challenges in Image Processing, Semantic Segmentation, Object Detection, Classification, and Tracking from up to 2 kilometers away, in real-time, in various operating environments, and in all weather and lighting conditions, Sign and Signal Detection, Anomaly Detection, range and kinematics estimation, SLAM for Image-based Navigation and GIS Mapping, Automated Emergency Braking, Predictive Maintenance, and others.

Rail Vision Ltd

Algorithm Group Manager

L4B Automotive (a division of L4B Software) offers end-to-end solutions (from POC to SOP), enabling rapid innovation to run efficiently on production-grade HW. The company, headquartered in Germany, has two branches there, one in Israel, and one in Romania.During my tenure, L4B has transformed from focusing on the Israeli start-up Embedded Linux market to rise through the ranks of the German Automotive Software industry. We have been working with some of the largest car manufacturers in the VDA, and various of their suppliers, on multiple projects for their upcoming production vehicles. Some of these multi-million dollar projects involved products in FCW ADAS,  IVI in the form of MGUs & ICs, DM & UCM, virtualization and MCAL in ECUs.L4B is a trusted partner of NXP,  Renesas, and others. L4B supports its customers in all stages of PLM/ALM. Projects expanded from creating custom BSPs in OpenEmbedded Linux Yocto projects for ECUs or AOSPs for MGUs, to including high-level functions in SWC on top of an existing software framework (e.g. Vector and Elektrobit AUTOSAR Classic & Adaptive Platforms in MISRA C/C++, or GENIVI). Projects are typically implemented over 6 months to 2 years, and involve one or more cross-functional Scrum teams, through all SDLC stages.Since I joined, L4B opened two branches (in Stuttgart and in Romania), and has grown in headcount and diversity of personnel competencies and backgrounds. The group I lead currently counts 18 in total.L4B is certified ISO 9001, ISO 90003, and is in the process of TISAX and ASPICE certification.As part of my contribution, we published "Detect distances effectively:Disparity image segmentation for driver assistance systems", Elektronik automotive, vol. 6 (2019).L4B is an active member of the Open Source community, and is looking forward to continue making more substantial contributions.

L4B Software

R&D Group Lead

In charge of R&D of artificial perception, planning, and control systems which evolve on PCs and in the ​ cloud​ , in order to adapt them to running in constricted embedded systems.Researched, prototyped, specified, and led multiple internationally-distributed teams, in projects productionizing state-of-the-art algorithms:• ​ Offloading​ algorithms at the edge to dedicated coprocessors (CPU, GPU, VSP, ISP, DSP, etc.), designing a concurrent processing pipeline, and integration into production HW devices.• ​ Perception​ : Calibration and preprocessing ​various sources (multi-spectrum and depth cameras, multiple RaDARs, LiDAR, IMU, etc.), ​ fusion of multiple sensors & measurements​ and ​ State Estimation​ (3D Reconstruction, Object Detection, Video & RaDAR Tracking, etc.)• ​ Planning​ : ​ ​ Scene Understanding​ ​ for ​ ADAS Safety Functions (e.g. Forward Collision Warning for Euro NCAP Autonomous Emergency Braking)​ .Our research yielded a Region Growing and Merging method for (non-binary) disparity-image segmentation (Connected Component Labelling) for ADAS.Skills employed:Algos​ : Classical & Statistical (i.e. Deep Learning) Computer Vision • 3D Reconstruction • Video & RaDAR Tracking • Sensor FusionProgramming: OpenCV • Python • C++1x • Linux • Yocto

Head of Automotive Intelligence

I held the sole responsibility for a number of simultaneous,distributed, multidisciplinary projects in the fields of Edge Computing and Connected Devices (IoT), representing L4B from presales to closing.Some examples of past projects:● An algorithm pipeline for a revolutionary ADAS product, executing SOA CV algorithms at the edge● A semi-autonomous, distributed (multicasting), multi-sensor surveillance system● BSW modules & SWC for AUTOSAR Classic & Adaptive PlatformsResponsibilities:● Initiation​ : B2B presales, exploiting opportunities with strategic customers, responding to RfQs, and winning contracts● Process Integration​ : Leading multiple cross-functional teams, HW Suppliers, and Customer representatives (PM, System Engineering, HW, SW)● Scope​ : Requirement Elicitation & Discovery, assumptions & design constraints, analysis, deliverables, SMART KPIs, specification​, ​ traceability. Change Management and upselling● Risk​ : e.g. mitigation and re-assessment (e.g. FMEDA), ​ ISO 26262 up to ASIL-D● Budget​ : Procurement & effort estimation● Schedule​ : Prioritizing activities, timeboxing, ALM (e.g. versioned releases, integrations), SDLC (e.g. balancing V-Model & Agile), Project Plan (e.g. deliveries, workshops, milestones, allocation of shared resources)● Quality​ : ​ ​ ASPICE & ISO 15504, ISO 90003 & ISO 9001, SQA Plan, customer evaluation, Acceptance Review, and Change Requests● HR: Recruiting & technical interviews, on-the-job training & development● Communication: Stakeholder Management, expectations at various industries and tiers (including OEM), definition of customer goals, constraints, and priorities to ensure that the decisions of everyone involved are properly aligned. RASCI, reflecting progress in status meetings and OIL, open issues (implications and resolution proposals), escalation, project steering

Algorithm Team Lead

Responsibilities• Curated the company Data-Driven Value roadmap• Lead research & development projects• Sole Technical Ownership of all algorithms in production: OneRun offline recommender and SmartFeed Online API• Evangelizing M2M and AI!Online Products delivered:• Online Ad Serving Recommender Engine (for the NG Online-API): One-shot app-install prediction, using Logistic Regression and Field-aware Factorial Machine (FFM), and comparing with a Holt-Winters double-exponential smoothing approach• A/B testing framework. Analytically exact expression for Bayesian comparison of scaled binary processes outcomes, yielding 3 orders of magnitude speed-up vs. frequentist approach (​arxiv.org/abs/1709.03003​)Offline Products delivered:3. Prescriptive: Online Ad Serving Recommender Engines & API (SmartX, SDK including video ads), ​ Media Optimization Recommender Engine (OneRun MORE)​ , ​ Campaign-similarity-based Blacklist Recommender (OneRun), Mixed Incentivized (Fraud) Traffic Identification​ , Cold-Start Allocation Recommender (​ OneRun​ CSAR, pronounced "Caesar")2. Predictive: Online eCPM Optimization (SmartX), Campaign Capping OLAP1. Descriptive: Domain-modelling of our market space (“Publisher and Advertiser DNA Quantification”), in both quality and quantity dimensions. The resulting (normalized) database served as a building block for further algorithmsSkills employed:Frequentist & Bayesian statistical inference • point & interval parameter estimation • model evaluation & selectionRule-based classification • density-based clustering (DBSCAN) • Factorization Machines • GLMs • Logistic RegressionPython (SciPy stack, scikit-learn) • R • Scala (ScalaNLP stack) • spark.ml • Jupyter • R Studio • AWSJava 8 • Apache (Commons, Maven, Tomcat) • Google dev tools (Guava, Guice, Gson) • Guartz • Liquibase • Jukito • Logback •JavaScript • SuperAgent • Chai • AWS (S3, RDS, Redshift) • Redis • Jetbrains stack • ELK stack • Graphite stack • Atlassian stack • Slack

YouAppi

First Data Scientist

- Member of the founding development team of an algorithmic trading company.• Software engineering & development, employing object oriented design & programming (using C++ STL & boost), and multi-threading (using ACE, the Adaptive Communication Environment).• Innovation and participation in the process of system engineering & architecture design, all within an agile development environment.• In charge of some of the central system components as well as the integration of trading algorithms & strategies.• Layed the infrastructure for design & integration of novel quantitative finance algorithms & analytics (technical analysis).• Additional techniques: Template meta-programming, pre-processor macros.

NGGL Ltd.

Core Developer

- Instructing the Molecular Dynamics computer experiment, which introduces interested students to theoretical topics in analytical and statistical mechanics, stochastic processes, and numerical methods, and serves as an opportunity for first hand experience as a computational theoretician. The lab culminates in the students running a C program which performs an atomistic simulation of liquid Argon, and analyzing the results, producing structural and thermodynamic information.• My resposibilities include instructing the students in the background material, verifying their preparation for the experiment, accompanying them in performing the experiment, directing them in the final report requirements, and evaluating them along each step of the way.• Since the experimental materials were not updated since the lab's inception (many years ago), I also had to coagulate various sources and notes from previous years, along with new material I collected and edited (Simplectic integration, Monte Carlo integration, ergodic theory,   etc.)• <a href="http://www.tau.ac.il/~physchem/Dynamics/">http://www.tau.ac.il/~physchem/Dynamics/</a>

Tel Aviv University

Computational Chemical Physics Instructor

- I was the senior recitation master for the course "Thermodynamics", under Professor Haim Diamant (<a href="http://www.tau.ac.il/~hdiamant/teaching/2009/thermo/).%E2%80%A2">http://www.tau.ac.il/~hdiamant/teaching/2009/thermo/).•</a> As such, my responsibilities included delivering weekly two-hour frontal lectures, personally tutoring students, synchronizing the recitations with the Professor's lectures, maintaining the existing TA guide and producing new modules, and assisting the Professor with the preparation and administration of the final exam.• The relatively long time during which I held this position afforded me the opportunity to acquire and hone my teaching skills, as well as improving my fundamental understanding of the course material and related topics.

Thermodynamics Lead Teaching Assistant

- I was the lead lab instructor for the Computational Chemistry Laboratory, under Dr. Ephraim Eliav (<a href="http://www.tau.ac.il/~ephraim/complab.html).%E2%80%A2">http://www.tau.ac.il/~ephraim/complab.html).•</a> The lab is a graduate course meant to provide organic chemistry students the computational and quantum chemistry tools needed to accompany their synthetic research.• My responsibilities included delivering frontal lectures to the entire class, leading interactive teaching sessions, instructing lab groups (as well as grading their preparatory exercises and final reports), personally evaluating students within my lab group, developing lab exercises, maintaining the lab manual, and composing and grading quiz questions.• Since I began instructing in the lab, we have had multiple requests by research group heads and their students from other universities to have students enroll in our course, as they have found it uniquely adapted to their research needs.• Additionally, student demand had outrun lab accommodations three years in a row.• During the time in which I was an instructor in the Computational Chemistry Lab, I learned a great deal, both with regard to the management and leadership of an instruction team, and with regard to the course material itself (Linear Programming, Convex Optimization algorithms, Monte Carlo, Simulated Annealing etc.)• I tried to put an emphasis on the specific research interests of the students, and on how the Computational Chemistry Lab can give them the necessary tools for these applications.

Computational and Quantum Chemistry Lead Instructor

- My research ("Heat Conduction in Single Molecule Junctions", direct track Ph.D.) is a theoretical and numerical study of the transport of heat through Single Molecule Junctions (a single molecule interacting with bulk solids held at different temperatures).• The theoretical part employs non-equilibrium (Newtonian and quantum) statistical mechanics and thermodynamics, while the numerical part consists of stochastic molecular dynamics simulations.• The day-to-day work involves analytical calculations as well as developement of stochastic algorithms for molecular dynamics. Software developement and reverse engineering in C/C++ of open-source computational chemistry software, automation of data analysis with BASH and PBS (utilizing distributed/parallel computing), modeling and reproducible research in IPython (SciPy stack with Numba on PyPy), and calculations in MATLAB and Mathematica, all within a LINUX environment.• My research has implications in molecular electronics and other nanotechnologies, with such potential applications as heat removal in computer hardware and thermoelectric devices, as well as advancing our understanding of fundamental science.

Chemical Physics Ph.D. Research

Taught Physics and Chemistry to American high school students abroad, preparing them for the college advanced placement (AP) exam in those subjects.

AMHSI - Alexander Muss High School in Israel

Chemistry and AP Physics Teacher

- Chemical Kinetics of the Formation of Silver Nanowires in Thin Surfactant Solution Films"Methods and techniques:Synthesis: Self assembly of micelles and nanowires in thin film from colloidal Gold.Analysis: Surface plasmon IR, HR- and cryo-TEM, EDS.Contributed to "Formation of Gold-Silver Nanowires in Thin Surfactant Solution Films", Olga Krichevski, Einat Tirosh, and Gil Markovich, Langmuir; 2006; 22(3) pp 867 - 870; (Letter) DOI: 10.1021/la051618v

Physical Chemistry Summer Research Intern

Leadership of a combat unit (AOC 13A), as well as general staff experience (DIV AFSCOORD and FA RGT Assistant S3): Operational Planning, Training (FA BNs & RGT FTX, BDEs FSCE CPX), and Project Management.Currently Major in the reserves (AFSCOORD).

Inon Sharony