Lead development efforts using NLP and Machine Learning to classify and interpret unstructured business communications

Designed and implemented machine learning solutions for Hebrew language text processing, including differentiating between modern and rabbinic Hebrew, identifying genre and age of Rabbinic text. Made use of word and document embeddings, LSTM and CNN for language modeling, coding done in python using scikit-learn, tensorflow, word2vec, doc2vec, fasttext

Responsible for all technology strategy and development, focusing on machine learning for financial applications, transcription technologies and cloud technologies. The Voitrax platform captures voice, text, and other communication records and accurately organizes and stores them to ensure complete compliance with Dodd-Frank regulatory requests. The system uses a unique blend of advanced technologies to reconstruct swap trades. Our solution addresses the direct needs of Dodd-Frank 1.35a and 23.201-3.

I manage the newly formed Israel Development and Engineering Center at Barclay Capital. This is an offshore application development and engineering facility for Barclays, located in the Kiryat Atidim Park in Tel Aviv. Responsibilities included vetting Israeli startups for inclusion in Barclays Accelerator programs globally.

Engineering and IT Support Manager, Israel. Built an Engineering team in Israel for Barcap comprising UNIX, Network, Windows, Storage and Database engineers, working in conjunction with the UK and US based engineering teams. Was responsible for identifying Israeli startups for use by Engineering teams gloablly. Also was responsible for IT support for the Israel office, managing a team of three.

Market Data Engineering and New Technology Integration, NJ and Israel – Delivered low latency market data solutions (mostly using NYSE Technologies (Wombat) on Infinband networks). Managed the merger of legacy Lehman and “heritage” Barcap systems post Lehman takeover.

Managed team of 8 supporting all market data infrastructure in the Americas, including Wombat, Reuters, Bloomberg an numerous stand-alone systems.

-Manage and maintain Wombat based Direct Exchange Infrastructure (DEI). Infrastructure includes over 20 feeds (CTA, UTP, OPRA, ITCH, Arcabook, Arcabook for options, BATS, Direct Edge, NSX, Btrade, NYSE Openbook, NYSE Quotes, NYSE LRP, NOIVIEW, NYSE Alerts, Bpipe, CME, Eurex, Euronext, CBOT, RMDS Gateway); over 100 feed handler servers, and client APIs. Market Making, Algorithmic Trading, Program Trading applications receive market data from DEI. Specific responsibilities include:-Manage deployment of new feeds from circuit provisioning, software (feed handler) provisioning, configuration, testing, etc. - Manage “exchange driven changes”, from receipt of notifications, hardware/software upgrades, testing, notification of end user community, support of end user testing.- Manage performance/capacity upgrades- Troubleshoot infrastructure issues (latency, rvd retransmissions, feed handler bugs)- Market Data Engineering for new business requirements & technologies (e.g. collocation, infinband)- Assist client application developers with proper application design.

 Market Data Engineering. Project Manager for various market data infrastructure projects, including implementation of proprietary rvd monitoring system, deployment of TeamQuest monitoring platform to all market data servers, integration of market data specific monitoring into TeamQuest.

 Manage the Global Market Data Operations team (8 people). Responsible for all market data feeds, circuits, services and infrastructure supporting GMI (institutional business) in the Americas. Services include, RMDS, Bloomberg, NASDAQ, IDBs, ILX, NewsEdge, Bridge, Track Data, Reuters Dealing 3000, EBS, Reuters+, Mighter, MDM, MLGES, and Hawk. Team responsibilities include, maintenance and operations of above services and related hardware and software, optimization of above services, installation of new services, and second level help desk. Team operates under the Microsoft Operations Framework (MOF). Specific projects include regular capacity review, Hawk Deployment, Teleport data center build-out, consolidation of regional infrastructure into central locations, BCP planning.

 Manage the Global Market Data Applications team (5 people) Responsibilities include, Development and maintenance of proprietary Market Data applications, research and development of emerging Market Data technologies. Applications include: MDPS (Market Data Pricing Service), which manages the definition, execution, formatting and delivery of market batch extracts. Built with C++ and Perl on Solaris, fully redundant, with automatic failover where appropriate. System maintains real-time specific definitions and other metadata about the particular instruments in a feed, to allow very specific job definitions. Jobs include both snapshots and transaction logs. Output formats include standard ASCII as well as fixed length record mainframe data sets. Currently in production. RTP, a shared memory based cache to facilitate fast synchronous price queries. RTP handles all complex event driven APIs and error handling, facilitating rapid application development and deployment. Platforms supported are Triarch, TIB/RV. Used mostly by equity applications.Other applications include a variety of feed handlers, mainframe Market Data interfaces, market data entitlement systems, gateways between MDM (inventory) and entitlement systems. Served on the MDDL (Market Data Definition Language, an FISD/SIAA XML for market data) technical committee.

 Manage the development, and/or design, implement, deploy and maintain components of the Finesse system. Finesse is a proprietary Market Data distribution platform with over 1000 workstations, 60 servers, and dozens of applications, and served as the primary Market Data distribution platform for Lehman Brothers in the Americas. Components include vendor feed handlers, central servers, workstation daemons, client API’s (C/C++, Java, Corba IDL), Corba Server, ActiveX components, DDE server, DLL/XLL, Client front-end applications. All components support load-balancing and failover where appropriate. Feed handlers are written in C/C++ and run on Solaris and NT platforms. Feed handlers include Reuters/Triarch, Tib, Open Bloomberg, Bridge/Knight Ridder, RMJ, GARBAN, Cantor et. al. Clients API’s are available for Solaris, SunOS and NT in C/C++, Java, and Corba IDL. Also designed and implemented protocols for the transfer of page-based data, self-describing records, historical data and news headlines. Also introduced IP multicast based distribution of updates, and thread-safe market data APIs. Provide design/development support for application groups developing market data applications that interface with the Finesse system. This includes (but is not limited to) API usage support, proper application design to ensure robustness, integrity and reliability of data, failover and load balancing strategies, integration of Finesse applications with other non-finesse systems, development and delivery of training classes, composition of detailed documentation and sample code.

Designed, developed, implemented various components and enhancements to all UNIX versions of the MX market data application. Technologies used include TCP/IP sockets, UNIX IPC (shared memory), memory mapped files, proprietary messaging protocol, MOTIF, XLIB. Components developed and or enhanced include front end application (MX), workstation daemon, client API’s. MX ran on SUN workstations, IBM RS6000, Data General Aviion. All code written in C/C++. Also designed and implemented an early version of the current messaging protocol used in subsequent versions of Track Data software.

Designed and implemented software for statistical analysis of seasonal “spikes “ of SS7 network usage related to 800 number resolution. Composed formal report to owners/clients (Baby Bells) detailing our analysis and conclusions. Coordinated data collection and storage for the project. Software was developed in C and S (a statistical package).Tested various components of FACS (Facility Assignment and Control System). This system would designate various “loop” (outside the central office) facilities (phone lines et. al.) for the fulfillment of service orders. Specific components tested included proprietary C libraries for transaction processing in UNISYS and mainframe environments, as well as full system test with straight through processing of transactions.