Sakina Ouisrani

• ● Integrating AI across the private equity investment life cycle : Designed and deployed LLM-powered internal tools and RAG/agentic workflows, improving data extraction, enrichment and investment report generation for deal teams.

• ● Built evaluative frameworks for LLM quality assurance, including hallucination detection, structured extraction reliability, and performance monitoring in production.

• ● Built scalable pipelines integrating financial/alternative datasets with automated feature engineering.

• ● Engineered an automated comparables engine leveraging embeddings and vector search for real-time

competitors benchmarking. Key Technologies: Python, OpenAI/Anthropic APIs, Langgraph/fuse, Llama-parse, vector search, embeddings, RAG, Databricks, DBT.

• ● Led development of SESAMm's ESG controversy detection engine. From PoC prototype to covering 200K+ public/private companies while processing 19B+ articles data lake across 6 languages. Fine-tuned a FinBERT model for controversy classification. Was also actively involved in Sales

processes (presentations to prospects and clients), recruitments and AI roadmap design.

• ● Pioneered the use of LLMs to improve data quality and introduced human-in-the-loop labelling UI, optimizing precision/recall trade-offs while heavily reducing manual QA.

• ● Collaborated with Euronext to build NLP-driven indicators and scoring algorithms directly supporting the launch of 3 thematic indexes. Key Technologies: Python, AWS, Hugging Face, Pytorch, Scikit-learn, OpenAI api, CI/CD.

• ● Led 3 client PoCs with top-tier financial institutions (Milliman, CPPIB, Japanese AM). Implemented time series forecasting systems using machine learning and NLP features using SESAMm's alternative data feeds for various use-cases: ○ Quarterly sales growth forecasting on~1k North American companies. ○ S&P500 realised volatility forecasting using high-frequency volatility proxies.

• ● Contributed to SESAMm's proprietary Python library for quantitative Machine Learning. Developed features in modules (feature extraction, temporal cross-validation, model interpretation) Key Technologies: Python, NumPy, Sickit learn, XGBoost, SHAP, time-series analysis.