05/18/2026 updated
40 % partially available
Senior AI Engineer | Agentic LLM Systems | AWS Bedrock & Azure OpenAI
Pune, India
Worldwide
Master of Science in Data Science, Lancaster University (Distinction)GenAILLMBedrockArtificial IntelligenceCloud ComputingPython (Programming Language)Machine LearningPresalesSolution ArchitecturePrompt EngineeringApi DesignApi GatewayDatabricks
Agentic LLM System Design
Expertise in designing and shipping production-grade agentic LLM systems using Python, LangChain, LangGraph, Pydantic AI, OpenAI Agents SDK, Anthropic Agents SDK, FastMCP, and MCP tool design, including multi-agent orchestration, RAG, semantic caching, prompt engineering, and skill-routed agent patterns.
Cloud Architecture on AWS and Azure
Deep hands-on experience with AWS (Bedrock, Lambda, API Gateway, S3, EC2, ECS, Cognito, Secrets Manager, Glue) and Azure (AI Foundry, OpenAI Service, App Insights, DevOps, Web Apps), as well as Databricks and GCP for end-to-end enterprise AI deployments.
Retrieval and Vector Store Integration
Proficiency in integrating retrieval and vector store solutions, including PostgreSQL with pgvector, Pinecone, ChromaDB, Qdrant, and Redis for semantic caching, enabling high-performance RAG pipelines in production environments.
LLM Providers and Inference
Practical experience with multiple LLM providers and inference frameworks, including AWS Bedrock, Azure OpenAI, Anthropic Claude, Hugging Face Transformers, Ollama, and vLLM.
Backend and API Integration
Strong backend development skills using FastAPI, FastMCP, OpenAPI, Celery, RabbitMQ, and Apache Airflow for building scalable, production-ready AI service backends.
DevOps and CI/CD
Hands-on experience with DevOps tooling, including Docker, Jenkins, Azure DevOps CI/CD, Ansible, Git, and GitHub Actions for automated deployment pipelines.
Observability and Monitoring
Implementation of observability solutions using MLflow, OpenTelemetry, and Azure Application Insights for tracking model performance, latency, and cost in production.
Traditional Machine Learning
Solid foundation in traditional ML using Pandas, NumPy, scikit-learn, PyTorch, TensorFlow, and Nixtla for time-series forecasting, fraud detection, and computer vision tasks.
Computer Vision and Fault Detection
Built a fault-detection system using computer vision and pattern matching to identify silent faults in PLC programming, achieving 96% efficiency improvement with reduced assembly-line disruption.
GPU-Accelerated AI Pipelines
Engineered distributed, GPU-accelerated AI/ML solutions processing 200,000+ documents per day, reducing per-document inference time by approximately 78% compared to CPU baseline.
Experience
Professional experience0 - 5 years
Freelancer since2026
Languages
EnglishFluentFrenchBasic knowledge
Project history
Forward-deployed operating model with ownership of customer discovery, architecture, rapid POC-to-production delivery, and handoff for enterprise GenAI engagements. Led a seven-person cross-functional AI lab. Key projects include an EU flight-catering client ServiceNow multi-agent system, EDI 835 Medicare claim-payment automation on AWS Bedrock, a SQL-to-Databricks AI migration agent, and a 10-track US lending and mortgage AI accelerator programme.
End-to-end AI engineering across logistics, banking, and manufacturing. Delivered air cargo tonnage forecasting models, LLM-powered IVR for logistics, a logistics chatbot for airport and dock management, a GPU-accelerated document AI pipeline, high-volume FASTag banking infrastructure, and a computer-vision fault detection system for automotive PLC assembly lines. Established LLM technical leadership and conducted training for 25+ employees.
Built an end-to-end ML solution for bonus-abuse fraud detection on approximately 100,000 monthly transactions, achieving 89-95% accuracy in production, deployed with Docker and Apache Airflow. Created custom prediction-threshold optimisation aligning model decisions with business priorities, projecting accuracy lift to 96-98%.