Banking Sector • Led the development of LLM-based solutions using LangChain and Hugging Face, including prompt-tuned models for document summarization and anomaly detection. • Developed multimodal and graph-enhanced RAG pipelines to support complex banking documents (e.g., loan contracts, account statements), using ColPali, VisualBERT, and Graph Neural Networks (GNNs) to enhance retrieval and generation accuracy. Implemented multivector retrieval using ColBERT to improve document matching across regulatory texts and policy documents, achieving higher recall in compliance-related queries. • Optimized retrieval pipelines using FAISS and hybrid search (BM25 + DPR) for low-latency access to financial documents, improving response relevance in customer-facing GenAI systems. • Architected and deployed GenAI systems for corporate transaction reconciliation, integrating Amazon SageMaker, Lambda, and API Gateway with RAG models to generate explanations for mismatches and automate resolution workflows. • Built cloud-native ML pipelines with Vertex AI, including training, tuning, and deploying models at scale via Kubernetes and CI/CD workflows. • Delivered rapid AI POCs by translating business goals into technical blueprints, collaborating with cross-functional teams across product and data science. • Mentored junior engineers in RAG implementation, GenAI tooling (LangChain, Transformers), and containerized deployments using Docker and Kubernetes. • Develop and maintain CICD pipelines for ML workflows. Developed a Generative AI-powered system to automate the reconciliation of corporate transactions across multiple accounts and systems. • Built evaluation workflows for RAG models, using Recall@K, BLEU, and ROUGE metrics to benchmark retriever and generator performance over real-world banking datasets. • Fine-tuned RAG models on domain-specific corpora, including internal banking policies, product descriptions, and transactional logs, enabling accurate generative QA and summarization capabilities. • Deployed a generative text summarization model using Amazon SageMaker. • Integrated with AWS Lambda and API Gateway for serverless access. • Used Vertex AI Workbench and Duet AI to enhance coding, documentation, and data exploration. • Deployed ML models to Vertex AI Endpoints, enabling secure, scalable predictions with monitoring. • Implement machine learning algorithms and models. Analyze large datasets to uncover trends and insights. Leverage predictive and AIML modeling techniques. • Built ML pipelines using Vertex AI Pipelines for model training, evaluation, deployment, and monitoring. • Implement Camunda orchestrator on Kubernetes