Santiago Daleffe

I build production machine learning, RAG, and event-driven systems for B2B products, turning predictive models and data pipelines into stable, scalable software.

On Fuccina Suite, I built Slancio’s real-time abandoned-cart recovery engine with dynamic margin-based coupons, and developed Egida’s fault-tolerant dunning system for failed payment recovery with exactly-once processing.

For Bussola, I developed a quantitative trading engine that processes macroeconomic news and OHLCV data, achieving a 0.81 Sharpe Ratio for next-day ETF closing-price predictions. I also implemented an LLM-based risk manager for auditing signals and simulating portfolio operations in an isolated AST Sandbox.

I’ve led ETL work that transformed unstructured legal records for Power BI analytics, evaluated LLM hallucinations and edge cases through HITL workflows, and contributed to Whoosh by modernizing its Python 2 codebase to Python 3.