Gattu Uday

Over 3+ years of experience architecting and delivering scalable AI/ML solutions across AWS and GCP, with deep specialization in Generative AI, LLMs, and transformer models like GPT-4, BERT, and T5. Designed and deployed full-stack GenAI systems for chatbots, document summarization, code generation, and synthetic data using HuggingFace, TensorFlow, and Amazon SageMaker. Built modular RAG pipelines with FAISS, Pinecone, Weaviate, OpenAI, and Cohere APIs, incorporating prompt chaining, vector search, and instruction/prompt tuning via LangChain and PromptLayer.

Engineered containerized LLM microservices using Docker and Kubernetes, deployed to AWS EKS, with auto-scaling via CloudWatch and infrastructure as code using Terraform and CloudFormation. Developed secure, real-time GenAI apps with React, Flask, Slack, and Teams integrations, enforcing authentication via AWS IAM/Cognito and ensuring enterprise-grade data compliance. Integrated RESTful APIs with AWS Lambda, EC2, and Step Functions to orchestrate inference and retraining pipelines, including multi-stage model evaluation and post-deployment workflows.

Applied MLOps best practices using MLflow, DVC, GitHub Actions, and CodePipeline for CI/CD automation, model governance, reproducibility, and cloud-native experiment tracking. Performed real-time data transformation and analytics using Amazon Glue, Athena, and GCP BigQuery; implemented feature engineering pipelines for model training and experimentation. Proficient in Python, SQL, JavaScript, and Bash for GenAI orchestration, pipeline automation, backend development, and API scripting across microservices and serverless stacks.

Led GenAI adoption across financial and cybersecurity domains, reducing inference costs by 38% using opensource LLMs, and driving innovation through PoCs, model evaluations (BLEU, ROUGE, perplexity), and research on diffusion models and responsible AI.

Generative AI Engineer at Capital One (2024-10 – Present)

• Designed and deployed LLM-powered assistants using OpenAI APIs, FAISS-based RAG, and AWS Lambda to streamline internal documentation access and improve developer productivity.
• Built multi-turn conversational agents for financial advisors using HuggingFace Transformers and LangChain, integrated with enterprise APIs for real-time customer and transaction insights.
• Fine-tuned GPT-4 models with dynamic prompts for financial use cases like transaction analysis, deployed securely using Flask, Cognito, and compliance-aligned access controls.
• Engineered GenAI solutions for multimodal tasks such as image captioning and PDF summarization using SageMaker Pipelines, Step Functions, and Pinecone vector embeddings.
• Deployed containerized LLM services to AWS EKS using Terraform and implemented auto-scaling with CloudWatch metrics for performance consistency under variable workloads.
• Built proprietary GPT-4-based systems for secure document Q&A on regulatory filings and a Python code refactoring assistant that boosted development speed and maintainability.
• Created enterprise copilots using LangChain agents for automated reporting, policy lookup, and analytics, increasing operational efficiency across business units.
• Implemented PromptLayer-based A/B testing with telemetry stored in DynamoDB to refine prompts and improve consistency across production GenAI applications.
• Reduced LLM inference costs by 38% by strategically replacing GPT-4 with AWS Bedrock and open-source models like Falcon, maintaining throughput and accuracy.
• Developed LLM evaluation pipelines using BLEU, ROUGE, and retrieval metrics, visualized in Grafana, while enabling secure internal experimentation with SageMaker Studio sandboxes.

AI/ML Engineer at Target (2021-08 – 2023-12)

• Developed scalable ETL pipelines using PySpark and AWS Glue to process e-commerce clickstream data for behavior prediction and customer analytics.
• Built and deployed recommendation engines using collaborative filtering models on SageMaker endpoints, supporting A/B testing and personalized product rankings.
• Engineered real-time analytics pipelines using Kinesis Data Streams, Lambda, and Redshift to optimize inventory movement and supply chain operations.
• Automated feature extraction workflows from product metadata using Python and NLP models, enhancing pricing optimization engines with enriched attributes.
• Designed and orchestrated ML workflows using Apache Airflow and SageMaker Pipelines, integrating CloudWatch for drift detection and automated model retraining.
• Migrated legacy batch ML jobs to serverless architectures with AWS Step Functions and Lambda, reducing latency by 50% and improving operational efficiency.
• Built customer segmentation models using clustering techniques and delivered interactive visualizations via Amazon QuickSight for marketing and merchandising teams.
• Integrated ML APIs into React-based internal tools, enabling real-time insights delivery for business users in logistics, pricing, and product planning teams.
• Led CI/CD pipeline development using GitHub Actions and Terraform, managing deployments of ML models and infrastructure across multi-region AWS environments.
• Implemented model governance with S3 versioning, metadata tracking, rollback mechanisms, and ethical documentation through model cards and fairness metrics.

Masters in Information Technology – Lindsey Wilson University