A highly passionate and results-driven AI Engineer specializing in Generative AI, Hybrid RAG, and Agentic AI Systems.
With 1.8 years of hands-on experience across the full AI project lifecycle, I design, build, optimize, and deploy scalable AI-driven solutions for real-world use cases.

• Agentic AI & Orchestration: CrewAI, LangChain Agents, LangGraph (stateful & multi-step agents)
• RAG Architectures: Hybrid RAG (Dense + Sparse), LightRAG, Traditional RAG Pipelines
• Search & Retrieval: Elasticsearch (BM25 + Vector Search), Hybrid Semantic Search
• LLM Inference & Optimization: Llama.cpp Server, vLLM, Ollama
• Generative AI Applications: Conversational AI, Knowledge Assistants, Autonomous & Tool-Using Agents
• Fine-tuning & Cloud Training: RunPod GPU Cloud, LoRA/QLoRA fine-tuning
• Deployment & Frontend: Streamlit Apps, Docker, API-driven integrations
• Cloud & Infra: AWS ECR, AWS ECS (Fargate/EC2), Task Definitions
• Programming Languages: Python (core), focused on AI frameworks

• 1.8+ years building and deploying LLM-powered systems in production environments.
• Designed and implemented Hybrid RAG pipelines combining Elasticsearch BM25 + vector embeddings for high-precision retrieval.
• Built LightRAG-based systems for low-latency, cost-efficient knowledge retrieval and reasoning.
• Developed LangGraph-based agents for stateful workflows, tool execution, memory handling, and complex multi-step reasoning.
• Architected and optimized RAG-powered chatbots, enterprise knowledge assistants, and multi-agent workflows using CrewAI, LangChain, and LangGraph.
• Optimized inference pipelines with Llama.cpp, vLLM, and Ollama to reduce latency and infrastructure cost.
• Built interactive Streamlit dashboards and conversational UIs for rapid experimentation and stakeholder demos.
• Implemented end-to-end AWS ECR + ECS deployments with secure secrets management, autoscaling, and ALB routing.
• Deployed and served fine-tuned models on RunPod GPU Cloud, leveraging LoRA/QLoRA for efficient training and inference.

• 🧠 LightRAG Contributor
  • Added a Google Gemini-based demo
  • Covers Naive, Local, Global, and Hybrid retrieval modes
  • Merged upstream into main branch
    🔗 HKUDS/LightRAG#2538