Sysmo leverages Machine Learning to predict stock-market volatility. By converting multiple types of Data (Internet Chatter, Technical Data, Context Data,...) into intensity...more indicators of the upcoming reaction of the market, we can anticipate price volatility on specific stocks and produce trading signals that can generate significant Alpha.
- Predictive models for unevenly sampled time-series with uncertainty quantification (Python). - Built an automated data pipeline: from raw data to automated cross-validation...more based feature selection to predictions. - Applied several Deep Learning approaches: CNN, LSTM, auto-encoders, transfer learning (keras, pytorch). - Set up collaborations with academic researchers at INRIA.
- Developed a Data-Driven causal model which explained the dependence of synaptic learning on the activity of neurons and experimental conditions. See https://github.com/iprokin/Cx-Str-STDP. -...more Worked with various experimental and synthetic datasets: Data Cleaning, Parsing, Transformation and Modeling (python). - Numerical Stochastic Simulations of Differential Equations, Parameter Optimization, Sensitivity Analysis. - 2 scientific publications (one in eLife, top 10% journal in neuroscience).
- Processing 64-dimensional time-series data recorded from neuronal cultures grown on multi-electrode arrays. - Developed a method for graph reconstruction from the time-series...more data generated by graph's nodes. - Time-series correlation and its statistical significance in C++ data manipulation/visualization in MATLAB.
2009 - 2013 (about 4 years)
- Solved numerically Differential Equations based model of a Neural Network with a customized Runge-Kutta in C++. - 2 international scientific publications describing the...more model of interacting neurons and an adaptive synapse.
Won a large project by solving a problem that our client considered hopeless; this brought 40k€+ / month. Built an automated data pipeline: from raw data to automated cross-validation based feature selection to predictions. Helped to win another very large contract by crafting custom model optimizing custom loss (we won the competition with two other startups).