Artificial intelligence in trading. I was part of the trading team making ambitious efforts to build fully autonomous, AI-based, genetically optimized algorithmic trading system. I designed/redesigned and implemented several components of that system - Forecaster (integration between AI and the trading platform) and Indicator Library (highly concurrent system for real time financial indicator calculation).

I also designed and implemented persistence layer to significantly decreased trading system startup time from 25 minutes down to 15 seconds. In addition, I redesigned and implemented training data processing which reduced processing time of 10 year trading data from 72 to 24 hours.

Design and implement trading AI test container framework allowing to replay trading session for AI testing and debugging.

Trading system successfully launched and currently in production.

Team

Built out the enterprise team by finding, selecting and hiring 6 highly qualified developers and support engineers. Adopted agile software development process. Established code ownership principles, code release cycle, and test framework for writing automated tests. Organized and managed DevOps for all company enterprise products.

New Product : Structured Finance Analytics Web API

Designed and led implementation of web service that integrated advanced deal search, portfolio management, real time pricing and valuation of individual tranches and whole portfolios. Included CDO Monte Carlo stress testing analysis and recursive CDO^2 and ABSCDO deal analysis. This system was built to be OLTP fault tolerant and horizontally scalable by design, dynamically expanding and reallocating calculation resources based on load, rules, and process priorities.

Small Individual Projects

Log4c Designed and implemented log system (analog log4J) for C++ applications. The log system was accomplished with remote “log appender” implementation that enables remote debugging of any running C++ company instance worldwide.

Smart Cache. Designed and implemented distributed hierarchical cache system. Ran cache update daemon at every layer (database, middleware, client) that used timestamps to automatically invalidate the distributed cache across layers and exposed to clients via REST web service API.

Architected, designed, and led several software projects which became successful new company products. These include the Deal Engine - a high performance distributed CDO and ABS financial calculation service and the Deal Library - a RESTful web service to manage CDO and ABS offerings. These systems made the company’s primary revenue generating desktop product accessible as a web service. Cluster set up, dynamic job distribution, priority queueing systems, and failover design were all implemented in-house using rule engine (CLIPS) for permission system automation.

Stood up company’s first data warehouse (Oracle) hosting over 3 billion records, including numerous high scale daily ETL. Implemented server-side ( J2EE) and client-side web service (C++) for data retrieval and analysis. Created an internal QA system system built on top of the Deal Engine to automatically test new versions of calculation software by comparing cash flow.