Sila

Contributed to building a revenue growth management platform for a consumer goods client: worked within a large codebase to automate data pipelines processing heterogeneous sales, financial and promotional data, and created an LLM agentic layer enabling business users to analyse performance in natural language while surfacing opportunities and risk alerts.

Built an LGBM demand-forecasting model for a global appliance manufacturer's operations, consistently cutting WMAPE by 10+ pts (vs client’s prior model) across ~100 products.

Developed a framework for analysing financial time series through a trend-oriented lens, leveraging FARIMA–GARCH models to jointly capture long-memory dynamics and volatility clustering. Designed the modelling pipeline to ensure robustness to non-stationarity and heavy-tailed distributions.

Designed and implemented supervised learning models (Random Forest and Gradient Boosting) to forecast monthly ETF shareholder churn rate using market and proprietary features. Engineered the underlying database and conducted panel-data econometric analysis to capture cross-sectional heterogeneity and temporal dynamics across funds, optimizing signal extraction and reducing noise in large-scale datasets.

Built a signal-detection module across multiple cryptocurrencies and market/sentiment indexes (Fear & Greed Index, VIX, S&P 500), combining APIs with a NLP pipeline on financial news to detect momentum/trend signals, informing ETF rebalancing and anticipating asset moves to help ensure liquidity to buy when needed.