Senior AI Engineer at Accenture

Design and deploy LLM-powered agentic applications using RAG and LangChain/LangGraph for enterprise SaaS, healthcare, fintech, and AI-driven business intelligence clients. Built production RAG pipelines using pgvector over unstructured and structured data sources including EHRs, PDFs, and clinical notes. Architected LangGraph-based agents to improve response quality and latency using multi-step reasoning, parallel execution, and tool orchestration. Implemented MCP to enable reliable, structured tool usage within agent workflows. Used Claude Code to iterate on agent workflows and accelerate development. Fine-tuned open-source LLMs (LLaMA, Mistral) using LoRA/PEFT for domain-specific and sensitive data use cases. Developed custom evaluation and monitoring systems to track model performance, response quality, and system reliability. Built full-stack AI applications using React (frontend) and Flask (backend) for user-facing GenAI features. Collaborated with clients and cross-functional tea

Machine Learning Engineer at Amazon

Developed and deployed machine learning models for large-scale recommendation and personalization systems, improving user engagement and conversion. Built ranking models using deep learning and gradient boosting (XGBoost) on large-scale user interaction data. Designed and maintained data pipelines and feature engineering workflows using Apache Spark, SQL, and AWS (S3, EC2, SageMaker). Implemented real-time inference and batch prediction systems with low-latency models. Ran A/B experiments and performed model evaluation and tuning to improve performance and reliability. Optimized model inference performance via model compression and efficient serving strategies. Collaborated with cross-functional teams to productionize end-to-end ML pipelines including deployment, monitoring, and versioning.

Machine Learning Engineer at NVIDIA

Planned and deployed deep learning models for computer vision applications (image classification, object detection) using TensorFlow and GPU acceleration. Improved model training efficiency through CUDA optimization, mixed-precision training, and distributed training strategies. Built scalable data pipelines processing 1M+ images, increasing model accuracy by ~12% through advanced feature engineering and data augmentation. Optimized inference performance across GPU architectures, reducing latency in production environments. Collaborated with cross-functional teams to productionize ML models, integrating them into internal tools and workflows.