ishh056

Designed and deployed 4 production GenAI/ML systems — fraud detection, document intelligence, workflow analytics, and multilingual NLP — in Python using TensorFlow, PyTorch, LLMs, and RAG; achieved 60–80% reduction in processing time across all use cases. Fine-tuned LLM and NLP models with RAG and prompt engineering to automate regulatory query resolution across 500+ policy documents, reducing response time by 70% and improving retrieval precision by 40% through iterative model evaluation. Engineered a multi-engine OCR + CV pipeline with pixel-level forensics and anomaly detection processing 10,000+ documents; cut verification time by 60%. Deployed a multilingual platform supporting 100+ languages with role-based access for 3 user tiers, reducing operational turnaround by 80%; tracked experiments with MLflow. Gathered requirements from 8+ cross-functional stakeholders to define ML product specs and drive production delivery on schedule.