Sarita Vásquez

Designed and implemented a production-grade RAG-based QA pipeline integrating linguistic rules, curated knowledge bases, and constraint-driven prompts. Built multi-step prompt workflows instructing LLMs to act as precision validators rather than generators, significantly reducing false positives. Integrated LLM systems with memoQ via APIs and SDKs, ensuring compatibility with existing production tooling. Developed NLP components including NER-based anonymization (spaCy, Stanza), text validation, and classification logic. Performed systematic prompt and model debugging to improve reliability, scalability, and user trust.

Designed and deployed NLP and ML components within Machine Translation and language QA systems. Built and evaluated text classification, quality estimation, and prompt-based workflows for production use. Worked hands-on with engineers to integrate NLP models and prompt logic into live systems. Optimized system performance through metric-driven iteration and controlled experimentation. Enhanced LLM translation evaluation with prompt engineering.

Managed AI and NLP data projects supporting supervised and unsupervised learning models. Translated technical NLP requirements into delivery plans, timelines, and quality criteria. Coordinated cross-functional teams and ensured data readiness for production systems.