Senior AI/ML Researcher

Led a team of 3 researchers to design and implement AI/ML solutions for time-series forecasting in cryptocurrency and stock markets, focused on high-frequency trading.

Investigated and benchmarked advanced architectures including DisjointCNN, Temporal Fusion Transformer, and Informer, achieving the highest performance with Informer for multi-step forecasting (24 candles ahead).

Designed and experimented with custom loss functions (Custom R², Quantile Loss, GMADL, Balanced GMADL) to improve predictive accuracy and robustness.

Developed a modular experimentation framework for rapid training, evaluation, and deployment of models, including:

...