Samira El-Masri

Systematically decompose complex RAG queries into atomic sub-queries with source mapping, dependency ordering, and aggregation strategies.

Build citation extraction and verification for RAG with claim identification, passage matching, verification levels, and formatted output.

Role-based prompt engineering adopting Lead AI Engineer persona for expert technical analysis, risk identification, and practical guidance.

Build a hierarchical summarization system creating multi-level summaries from section to document level with coherence preservation.

Apply structured reflection to model selection decisions with assumption checking, alternative analysis, and trade-off evaluation.

Implement a comprehensive LLM retry strategy with error classification, exponential backoff, circuit breakers, and fallback chains.

Meta-prompt for systematic prompt improvement through structural, content, and optimization analysis with detailed change rationale.

Build adaptive chunking that analyzes documents and selects optimal chunking strategies based on content characteristics.

Apply chain-of-thought cost-benefit analysis for technology decisions with comprehensive cost, benefit, and risk evaluation.

Build an output length controller with estimation, prompt calibration, smart truncation, and expansion for precise length control.

Apply multiple expert personas to RAG optimization covering retrieval, NLP, systems engineering, and product perspectives.

Build a semantic search quality monitor with sampling, relevance evaluation, degradation detection, and comprehensive reporting.