Senior quantitative developer and data scientist in capital markets, specialising in LLM and agentic AI systems for trading workflows.

Experience across structured products, derivatives pricing and front-office data: equity and credit derivatives, autocalls and EMTN structuring, rates and FX, volatility surfaces, risk and sensitivities.

I build production-grade pricing and risk models, LLM and agentic pipelines that turn unstructured trader and client messages into validated structured data, and the data platforms behind them: validated against reference libraries such as QuantLib, tested, typed and run through CI.

Market finance (structured products, derivatives, credit, volatility) · LLM / agentic AI in production (extraction, RAG, evaluation, guardrails) · Machine learning for pricing and risk · Data engineering (Databricks, Spark, Delta Lake)

• autocall-pricer — Monte Carlo pricer for autocalls and worst-of autocalls (2–5 underlyings) under correlated Black-Scholes, with exact log-normal simulation, antithetic variates, bump-and-reprice greeks and correlation sensitivity; validated against QuantLib.
• credit-pricer — Credit derivatives library: ISDA-convention CDS bootstrapping, CS01, CDS indices, Gaussian-copula CDO tranches with base correlation, and CLNs; matches QuantLib's ISDA and midpoint engines to within 5e-4 bp.
• crypto-vol-surface — Arbitrage-free BTC/ETH implied volatility surfaces from Deribit's public API: raw SVI per expiry and SSVI with no-butterfly and no-calendar checks, smile analytics (ATM, 25Δ RR/BF), Black-76 pricing and greeks.
• market-data-lakehouse — Medallion lakehouse on Databricks for ECB FX, yield-curve and €STR data plus a synthetic tick stream: Delta Lake, Lakeflow pipelines and jobs, Unity Catalog governance, data contracts, quarantine and freshness SLAs, dev/prod via bundles.
• france-grid-analytics — French power system analytics on the ENTSO-E Transparency Platform API: 15-minute day-ahead prices, load, generation mix, cross-border flows and CO2 intensity, a walk-forward day-ahead price forecast that beats naive baselines (MAE −13% vs D-1, Diebold-Mariano tested), a European extension comparing France with five other bidding zones (price spreads, convergence, congestion on the NTC borders, negative prices vs solar), and a local Streamlit dashboard.

• Languages: Python · C# · SQL
• LLM / GenAI: LangGraph · LangChain · agentic workflows and tool calling · RAG (embeddings, vector stores) · structured output (Pydantic / JSON Schema) · prompt engineering · LLM evaluation and guardrails · fine-tuning (LoRA) · OpenAI / Azure OpenAI / Anthropic APIs · local models (Ollama, vLLM) · Hugging Face Transformers
• Data science / ML: pandas · NumPy · scikit-learn
• Data engineering: PySpark · Databricks (Delta Lake, Lakeflow, Unity Catalog) · FastAPI · Docker
• Quant: QuantLib · Monte Carlo · volatility surfaces · credit models
• Engineering: pytest · mypy · CI (GitHub Actions)

LinkedIn: youssef-el-bqaq