Created a research laboratory project for forecast creation.

Variety list of factors used for making non-linear time-series forecasts, using multi-density probability predictions.

The solution is searching for the most suitable factors, verify them, and create multi-level forecast core.

The solution for automated/semi-automated development of trading strategies for selected instruments, trading style (Momentum, mean-reverse, Stat-Arbitrage, scalping, ...) powered by deep learning, machine learning, genetic/evolution technics.

The solution include:

• Signal engineering engine
• Technical Indicator generator
• Forecasting module
• Backtesting section
• Walk-forward optimization
• Multi Cross validation optimization
• Probability of over optimisation module
• Transaction cost / Market Impact analyzer
• Robust testing
• Risk parameters analyzer
• Portfolio section
• Synthetic ETF generator
• Data module

Developed multi-server solution for automated trading strategy generation.

Key Features:

• Multi-Server scalable solution.
• A genetic algorithm for a finding of strategy architecture.
• In house non-convex numeric optimizer for strategy and portfolio optimization.
• Trading strategy portfolio creation.
• In-house signal based, fast backtesting engine.

The solution for creating, rebalancing, testing the portfolios.
Working with a different type of assets including alphas.

Some of the features:

• Implied various tools for creation portfolio (CAPM, APT, Black-Litterman, ...)
• Using various tools for forecasting of assets (ARMA, GARCH, Kalman filters, MCMC, Deep-NN, LSTM, ... )
• Factor designing module.
• Stress testing / Robust module.
• Walk-forward simulation, What-if testing, ...
• Portfolio with active signals testing

The system for real-time analysis of all trades from MT4-Manager in order to find suspicious or risky client behavior.

In a phase of developing a machine-learning scoring system was found a set of clients, that trades extremely well, but for a limited amount of time, and then disappear. Designed advanced systems for finding that kind of traders, combining their trades into a portfolio with applied risk metrics and sending that trades to any MT4 terminal.

Practiced machine learning techniques for analyzing and clustering of 200 000+ clients. Data includes registration information, trading activity, transactions.

Applied research results for a variety of practical applications with the aim to maximize customer lifecycle included:

• helping to target the clients with relevant trading ideas and suggesting new trading opportunities,
• profiling clients behavior and motivations,
• analysis of trading style, customizing education plot and personal guidance.

Due to applying of a developed tool for dealing, equity profitability increased by 50+%. Obviously, it is a result of a set of actions, where an only part was developed tools for dealing.

Few examples:

• The solution for risk monitoring, forecasting of exposure and customer profitability.
• The system for early monitoring of suspicious client trading activity, and real-time generating of advanced client report.
• Updated way of spread markup operation algorithm from random to the advanced analytical system, which allows keeping average spreads at a certain level, while effective spread increased by 2-10%.

Implement various solution, using Keras(TF) and PyTorch for Deep Reinforcement learning, forecasting, designing trading strategies, trading strategies market module.

As well apply NN framework for various tasks, including price forecasting, risk analyses, exposure forecasting, artificial data generation, ...

The solution for extensive analyses of trade strategies and conducting details reports. Bring insights of trading weak points, risk estimation, a potential for improvements.

Include 70+ analytical module, which can be custom used by a user, depending on a type of analyses needed.

One of the crucial parts of the analytical system is a cutting edge forecasting technology, that was made especially for trading systems and brings a dramatic increase in accuracy of risk estimations.

Another one is a multiple mechanism of verification trading strategies, in order to discover hidden risks, and mesure the robustness level.