Tejaswi Reddy Vallapu

Machine Learning Engineer at JP Morgan Chase & Co. (2025-07 – Present)

Deployed AI-driven financial analytics on Azure ML and Databricks using transformers, RAG, and Azure OpenAI to automate portfolio and compliance insights, cutting review time by 60%, saving 15+ hours weekly per analyst
Built fraud-detection and credit-risk models using XGBoost, PyTorch Lightning, and Azure AutoML, improving precision by 25% while embedding explainable AI components - SHAP and LIME for enhanced audit transparency
Implemented containerized ML microservices using Docker and Kubernetes on Azure, integrated with Event Hubs and REST APIs for real-time model inference, achieving 89% uptime in production
Engineered ETL pipelines via Azure Data Factory and Delta Lake to process 50M+ financial records daily, ensuring high data reliability, lineage tracking, and end-to-end automation across analytics workflows
Designed a robust RAG-based Q&A chatbot leveraging LangChain and Gradio by integrating LLMs with a vector database, enabling natural-language querying across 200+ complex datasets
Achieved a 60% boost in query-resolution speed and a 50% gain in retrieval precision, saving the data team 10+ hours weekly using Chain-of-Thought and Few-Shot Prompting techniques

Machine Learning Engineer Intern at MetLife (2024-12 – 2025-04)

Designed a constrained risk-scoring framework that decouples severity and frequency models for insurance claims prediction, enabling formal actuarial compliance guarantees across underwriting pipelines
Integrated Bayesian uncertainty quantification with Student's t-based confidence intervals to enforce probabilistic reliability thresholds and provide formal performance guarantees during model validation and evaluation
Fine-tuned Gradient Boosting and deep neural network ensembles for claims triage and fraud scoring, improving inference reliability by 15%; deployed reproducible ML pipelines with auditable validation tests for insurance compliance
Automated preprocessing pipelines including covering feature engineering, data validation, and train-test preparation across 2 datasets, cutting preprocessing effort by 35%
Deployed lightweight FastAPI + Docker inference pipelines for internal ML model serving, eliminating manual model handoffs and reducing validation time by 40%

ML Engineer / Associate Software Engineer at Accenture (2023-01 – 2023-06)

Enhanced sales-to-forecast alignment in the Snowflake data warehouse using multi-objective loss optimization and feature normalization to improve downstream tasks such as pricing prediction, demand forecasting, and revenue analytics
Explored 10+ feature engineering variants using contrastive learning schemes, achieving 6.2% improvement in forecast accuracy and 9.6% gain in model recall, setting new performance benchmarks
Built a Business Intelligence Summariser leveraging structured sales and finance data to project it to a unified analytics space, generating concise revenue summaries and providing insightful responses to 10k+ business queries

Machine Learning Engineer at LTI Mindtree (2020-12 – 2022-12)

Led a 3-member engineering team to architect and deploy a geo-distributed climate analytics system from scratch, integrating multimodal deep learning across edge, HPC, and cloud, delivered end-to-end within 6 months
Reduced model training latency by 35% through GPU acceleration (CUDA), distributed TensorFlow pipelines, and AWS SageMaker, enhancing prediction throughput by 20% and compute efficiency by 15%
Incorporated Federated Learning with lightweight ML algorithms (XGBoost) on low-resource edge nodes, achieving 97% rainfall prediction accuracy while ensuring data privacy and scalable MLOps workflows
Engineered multi-variable time-series forecasting pipelines using LSTM and PyTorch, processing 100+ years of oceanographic and weather data across 10+ climate variables, improving long-horizon prediction stability by 22% and setting new evaluation benchmarks across regional climate datasets

About