Hanna Gupta

Developed ML pipeline processing 500K+ NHS patient records with temporal/weather data, improving resource allocation by 35%. Enhanced model with SMOTE sampling and time-based validation splits, achieving 95% improvement in no-show predictions. Implemented ensemble models (XGBoost/Random Forest), reducing features by 40% while maintaining 95% accuracy and cutting costs by 30%. Built FastAPI server for FOI/SAR document intake leveraging LLMs for 85% faster contextual multi-class classification with 99.9% uptime. Developed transformer and graph-based entity detection system with Azure PII and NER, improving redaction accuracy from 65% to 90%. Engineered Azure/PostgreSQL document pipeline with fine-tuned LLMs, reducing processing time by 97% (from 5 days to 15 minutes) while maintaining 100% compliance. Implemented RAG pipeline with vector database for automatic answer retrieval.

Led 6+ Generative AI solutions for the UK public sector, including social services automation that reduced documentation time from 4 hours to 45 minutes for 200+ workers (80% efficiency gain). Built OpenAI/Flask-based request triage system handling 5K+ weekly requests with 95% accuracy and CI/CD deployment reducing response time by 60%. Implemented RAG system with Pinecone/LangChain achieving 99.9% data security while reducing retrieval time by 70%. Created multimodal (LLM and diffusion model) comic generator producing 1,000+ educational materials with 85% satisfaction.