Hire Bharath P. - LLM Engineer

* Build and deploy LLM-powered applications using RAG architecture with FAISS and Pinecone vector stores, enabling automated document review and compliance checks across global banking operations. * Develop NLP pipelines for transaction classification and fraud narrative analysis using fine-tuned BERT and RoBERTa models served through AWS SageMaker endpoints. * Maintain scalable ML inference services on AWS SageMaker and EKS, handling real-time prediction * workloads with low-latency requirements across payment processing systems. * Apply prompt engineering techniques including Chain-of-Thought to improve LLM accuracy on regulatory document summarization and internal knowledge extraction tasks. * Manage MLOps workflows using Docker, Kubernetes, and CI/CD pipelines (Jenkins, GitHub Actions) for * model versioning, testing, and production deployment. * Fine-tune transformer models (BERT, T5) on financial domain text for entity extraction and document * classification, iterating performance through systematic evaluation. * Work closely with product managers, data engineers, and compliance teams to scope AI features and * translate business needs into technical solutions. * Track model performance in production using MLflow and CloudWatch, setting up drift detection alerts and retraining triggers.

* Developed NLP models for customer interaction analysis using BERT-based architectures for intent * classification and sentiment scoring on transaction and support data. * Prototyped an internal question-answering tool using OpenAI APIs and FAISS for retrieving relevant sections from policy documents and regulatory filings, validating feasibility for enterprise knowledge retrieval. * Built feature extraction and data preprocessing pipelines using PySpark and Apache Spark to handle * unstructured text data at scale, cutting down model training time significantly. * Set up model monitoring using MLflow and CloudWatch to track prediction quality and trigger alerts when performance dropped below acceptable thresholds. * Developed recommendation models using collaborative filtering and neural network approaches to surface relevant financial products to customers. * Ran A/B tests to validate model improvements before production rollout, working with analytics teams to ensure statistical rigor in results. * Processed and analyzed large structured and unstructured datasets using SQL, Pandas, and PySpark to * support trade finance decision-making.

* Contributed to the migration of legacy Hadoop-based data infrastructure to Snowflake, writing data * transformation scripts in PySpark and Snowflake SQL to convert existing Hive queries and MapReduce jobs. * Built NLP pipelines using spaCy and TensorFlow for extracting and classifying information from regulatory filings, utility compliance reports, and customer correspondence. * Developed ML models on AWS SageMaker for energy demand forecasting and outage prediction using * historical grid and weather data, helping improve forecast accuracy over legacy rule-based approaches. * Write ETL and data transformation pipelines using PySpark, SQL, and Snowflake SQL to restructure and * move datasets from HDFS into Snowflake tables. * Built ML serving endpoints using Flask, containerized with Docker, and deployed on AWS for consumption by internal analytics and operations teams. * Created data ingestion pipelines using Kafka and PySpark for near-real-time processing of energy meter readings and grid sensor data. * Performed data validation and reconciliation checks during the platform migration to ensure consistency across hundreds of migrated datasets. * Worked in an Agile/Scrum team alongside data engineers, business analysts, and SCE stakeholders to plan sprints and deliver project milestones.