In the making!
Developed pipeline for generation of metabolic graphs for human gut microorganisms in Python. Built predictive model for cooccurrence and compatibility of bacteria pairs using graph embeddings, achieving 73% accuracy in the models overall. Achieved 98% out-of-sample accuracy for predicting vitamin biosynthesis pathways using LASSO regression. Investigated research literature on biocomputational methods for microbial community analysis.
Automated custom querying of a REST API used for market data, saving upfront work of up to 5 hours per project. Built a parser for XML query responses to query market data critical for real-time trading and pricing of ~200 assets. Improved library calculating daily and minutely volatility estimators using calculations in academic literature such as Garman-Klass estimation, using kdb+ queries to make the querying and estimation 10x faster.
Adapted existing market data pipelines to new stock exchanges, and added filters for instrument types. Performed data cleaning, linear regression, correlation analysis and signal processing on market data using Python. Implemented a covariance estimator for high-frequency, noisy and asynchronous data (project won 1st place at intern hackathon).
Identified higher level structures in logical descriptions of games using Java and Answer Set Programming. Improved the space complexity of a Monte Carlo Tree Search algorithm used for AI game playing agents. Worked with convoluted Java codebase and tackled technical debt by increasing the code documentation level to 90% and modularizing the architecture.