Anan Ashraf

Engineered prompt guidelines for QC teams to generate culturally grounded datasets; supported RLHF preference ranking; conducted R&D on post-training data strategies and LLM-as-a-Judge evaluation frameworks; localized benchmarks MT-Bench, XNLI, IFEval, XCOPA into Southeast Asian languages.

Deployed and automated CI/CD production ML pipelines on GCP Vertex AI using Cloud Build for pricing and promotion optimization for a Fortune 500 client; implemented data quality validation and monitoring across enterprise datasets; extended internal GenAI platform by integrating MySQL, MariaDB and Postgres for structured data reasoning and a conversational memory; built GenAI proofs-of-concept for BFSI use cases including financial chatbot systems and automated narrative generation pipelines.

Engineered core components on an internal GenAI platform (Python, LangChain, HuggingFace) applying LLMs to structured enterprise datasets; benchmarked performance; scaled a lumber production planning system using XGBoost and Random Forest on AWS SageMaker; automated data pipelines reducing manual effort by 85%; improved NLP duplicate-invoice detection; built VAT engine and tax forecasting; consulted for international Fortune 500 clients; designed a Streamlit-based resume ranking system.