Hao V. P.

🚀 Expert AI Agentic Engineer | LLMs | RAG | Context Engineering | Agent Platform I help companies turn AI ideas into working, production-ready systems - from RAG and knowledge graphs to multi-agent workflows, AI assistants, ML pipelines, APIs, and cloud deployment. ✅ I build AI systems for real business environments. My project experience spans several complex domains: ᨞ Legal AI: GraphRAG, hierarchical legal-document retrieval, citation verification, agent workflows and knowledge graphs. 🩺 Healthcare / Recruitment AI: clinician and job-search assistants, structured tool calling, conversational memory, business-rule validation and API integration. 🧑🏻‍💻 Enterprise IT / Service Management: AI-assisted incident classification and assignment using historical organizational knowledge, vector retrieval and human accept/override workflows. 🔍 Knowledge & Research Systems: document ingestion, semantic search, graph algorithms, agent memory and evidence-grounded question answering. ✅ What I bring to a project 1️⃣ End-to-end AI engineering - Design and implement AI systems from architecture to deployment. - Experience across FastAPI, LangGraph, Neo4j, PostgreSQL/pgvector, Redis, Docker, Streamlit, Gradio, and cloud infrastructure. - Build APIs, streaming interfaces, persistent memory, retrieval pipelines, evaluation flows, and production deployment. 2️⃣ Advanced RAG & GraphRAG - Build systems combining vector search + graph traversal + metadata + hierarchical retrieval + LLM reasoning. - Hands-on experience with Neo4j AuraDB Professional, Cypher, graph retrieval, community detection, PageRank, centrality, shortest paths, Node2Vec, and Graph Data Science concepts. - Designed knowledge systems where answers are connected back to evidence, source documents, graph relationships, and citations rather than relying purely on model memory. 3️⃣ Agentic AI & LangGraph - Design stateful AI workflows with nodes, tools, routing, checkpoints, persistent state, sub-agents, streaming, and human-in-the-loop controls. - Designed architectures with 4 dedicated AI workflows and 3 specialized sub-agents operating over a shared knowledge corpus. - Build agents that can retrieve information, reason over evidence, call tools, validate outputs, and escalate uncertain decisions to humans. 4️⃣ Large-scale document & knowledge processing - Worked with document-processing pipelines covering 700+ PDF pages for embedding and knowledge retrieval workloads. - Build ingestion pipelines involving document parsing, chunking, metadata, embeddings, PII handling, graph construction, and retrieval. - Evaluate chunking, semantic retrieval, hybrid search, hierarchy expansion, and evidence aggregation rather than relying on a single retrieval strategy. 5️⃣ Multi-model / multi-provider AI architecture - Practical experience working across OpenAI, Groq, Google Gemini, Qwen, embedding models, and self/externally hosted model infrastructure. - Select models based on the actual workload instead of forcing one model to handle everything. - Track input tokens, output tokens, embedding usage, inference cost, latency, and model capability when making architecture decisions. 6️⃣ Machine Learning — beyond LLMs - Build and tune traditional ML models using LightGBM classification and LambdaRank learning-to-rank. - Experience with class imbalance, feature engineering, statistical testing, early stopping, hyperparameter tuning, model evaluation, and inference pipelines. - Comfortable moving between classical ML, graph ML, retrieval systems, and Generative AI depending on the business problem. ✅ How I approach AI projects I normally start with four questions: 1️⃣ What decision are we improving? 2️⃣ What evidence should the AI use? 3️⃣ What can safely be automated? 4️⃣ How do we measure whether the system is actually better? From there, I choose the architecture. That is the value I bring ➡️ I don't just integrate AI APIs, I engineer the system around the business problem. ✅ ✅ Tech stacks: • AI / LLM: OpenAI, Gemini, Groq, Qwen, LangChain, LangGraph • RAG: Vector Search, Hybrid Search, Semantic Retrieval, Reranking, GraphRAG • Knowledge Graph: Neo4j, Cypher, Neo4j GDS, Node2Vec, PageRank, Louvain/Leiden • Machine Learning: LightGBM, LambdaRank, Classification, Ranking, Feature Engineering • Backend: Python, FastAPI, AsyncIO, REST API, SSE • Databases: PostgreSQL, pgvector, Neo4j, Redis • Frontend: Streamlit, Gradio (for demo), Next.JS (for production) • Deployment: Docker, Docker Compose, Linux, Ubuntu, Cloud/VPS 😉 I'm always ready to connect. Please don't hesitate to message me.