Worked on solutions to enhance the decision-making and analysis for Traders and Liquidity Providers.Projects:- Developed LSTM time series models to forecast market volatility and spreads, improving algorithm performance.- Created predictive models for performance indicators in high-frequency algorithmic trading.- Built custom ranking and validation modules for decision-making processes.- Integrated models with the existing trading platform to prototype semi-automatic trading systems.- Experimented with Q-learning to optimize the selection of trade destinations.- Analyzed forex market impact and price decay using Hadoop and SQL-driven data analysis tools.

Designed and developed solutions using automation bots and APIs for the HOLMES AI platform.Projects:- Created a search system with synonym capabilities and a feedback-based recommendation feature.- Developed a hybrid classification system to understand user intent in a virtual assistant.- Applied LSTM models for automatic entity labeling in specific domains.- Designed a CNN system to classify application screens, automating IT support tasks.- Recognized for innovation, filed patents, and authored publications in NLP and Machine Learning.

DMSO solubility prediction using machine learning methods: - Worked on a tool for predicting DMSO solubility for a screening collection using Support Vector Machines and Bayesian Learning (Part of lead optimization, this model could enable vendors to better manage their compound libraries and reduce the cost of the drug development process). - Principal component analysis (PCA) to choose various molecular descriptors as input features and performed Time-series forecasting of descriptor changes.- Cross-validations and real-time validation (by experimental scientists) of the models helped in formulating a consensus using Decision Trees with standardized confidence levels to achieve least False negative predictions.- The model facilitated filtering out 30% of screening collection.