Sujan D.

𝗕𝗶𝗼𝗺𝗲𝗱𝗶𝗰𝗮𝗹 𝗼𝗻𝘁𝗼𝗹𝗼𝗴𝘆 & 𝐆𝐫𝐚𝐩𝐡-𝐑𝐀𝐆 𝐍𝐞𝐨𝟒𝐣 | 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲 𝗔𝗴𝗲𝗻𝘁 | 𝗣𝗶𝗽𝗲𝗰𝗮𝘁 | 𝐀𝐳𝐮𝐫𝐞 𝐒𝐞𝐚𝐫𝐜𝐡 | 𝗥𝗮𝗴 𝗙𝘂𝘀𝗶𝗼𝗻 | 𝗥𝗲-𝗿𝗮𝗻𝗸𝗶𝗻𝗴 | 𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰 𝗖𝗮𝗰𝗵𝗶𝗻𝗴 | 𝐓𝐨𝐆 (𝐓𝐡𝐢𝐧𝐤-𝐎𝐧-𝐆𝐫𝐚𝐩𝐡) | 𝗥𝗔𝗚 𝗘𝘃𝗮𝗹 | 𝐌𝐢𝐜𝐫𝐨𝐬𝐨𝐟𝐭 𝐆𝐫𝐚𝐩𝐡 𝐀𝐏𝐈 I specialize in developing RAG and AI Agent systems with 15+ years of full-stack experience using LangChain, LangGraph, FastAPI, and vector databases (Qdrant, Pinecone, FAISS etc). For data ingestion, I build end-to-end pipelines supporting multi-format sources (PDF, DOCX, CSV, Excel, websites), structured cleaning with spaCy, Textacy & CleanText, web scraping via Playwright, BeautifulSoup & Jina Reader, Overlap Late Chunking for retrieval accuracy, and Microsoft Graph API integration for SharePoint/OneDrive across Microsoft 365.I also build real-time Voice AI agents using Pipecat with full STT → LLM → TTS streaming pipelines. My retrieval expertise covers Re-ranking (Cohere, FlashRank, Jina), dense/sparse/hybrid search, RAG Fusion, Semantic Caching, contextual compression, custom output parsers, LLM Token tracing, and RAGAS-based evaluation. I am a 𝐍𝐞𝐨𝟒𝐣 𝐂𝐞𝐫𝐭𝐢𝐟𝐢𝐞𝐝 𝐏𝐫𝐨𝐟𝐞𝐬𝐬𝐢𝐨𝐧𝐚𝐥 specializing in Graph RAG, graph data modeling, Cypher queries, and domain ontology alignment. For graph ingestion, I have developed knowledge graph pipelines using GraphTransformers and NLP libraries for entity & relation extraction, aligned with domain-specific ontologies. For graph retrieval, I have implemented Graph Traversal, vector search, Think-On-Graph (ToG) reasoning, and Cypher Agents (generation, validation, correction, execution) leveraging LLMs for answers. Front End experience with TypeScript, Vue.js Composition API, Quasar Framework. I have a sharp eye for detail and quality and have experience of Software Reengineering. I am a self motivated individual who is both deadline focused and thorough in my approach to work. -------------------𝐏𝐫𝐨𝐯𝐞𝐧 𝐄𝐱𝐩𝐞𝐫𝐭𝐢𝐬𝐞 𝐢𝐧 𝐋𝐋𝐌 𝐀𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭------------------- 𝐀𝐝𝐯𝐚𝐧𝐜𝐞𝐝 𝐑𝐀𝐆 & 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 𝐒𝐲𝐬𝐭𝐞𝐦𝐬 • Implemented RAG with multiple data formats (PDF, DOCX, CSV, Excel, TXT, websites etc.) • Enhanced retrieval techniques using RAG Fusion with multiple re-rankers (Cohere, FlashRank, Jina) • Applied optimization strategies like Prompt Caching, Semantic Caching with vector databases • Context-aware chunking is implemented by "Overlap Late Chunking" to improve retrieval accuracy • Conducted RAG evaluations using RAGAS for quality assessment and benchmarking • Built structured web scraping and cleaning pipelines using Playwright, BeautifulSoup, Jina Reader, Textacy, CleanText, spaCy etc. • Implemented dense, sparse, and hybrid search for advanced similarity search and optimal retrieval • Implemented LLM Token Tracing for monitoring LLM cost per user • Implemented custom output parsers for specialized formatting • Integrated Microsoft Graph API to fetch and process data from SharePoint and OneDrive across Microsoft 365 𝐊𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞 𝐆𝐫𝐚𝐩𝐡 & 𝐆𝐫𝐚𝐩𝐡𝐑𝐀𝐆 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 • Built Neo4j GraphRAG system with Graph Traverse and vector search capabilities • Created knowledge graphs aligned with domain-specific ontologies including bio-medical Ontology, using transformers, custom logic and detailed instruction-based prompts. • Developed Cypher Agents with LangGraph for Cypher generation, validation, correction, and execution • MS GraphRAG solutions with LanceDB and Parquet files, search and Indexing ,incremental indexing. • Successfully migrated MS GraphRAG Parquet files to Neo4j (KG) database systems 𝐋𝐋𝐌 𝐀𝐠𝐞𝐧𝐭𝐬 & 𝐂𝐨𝐧𝐯𝐞𝐫𝐬𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐀𝐈 • Designed LangGraph Agent architectures for multi-step reasoning workflows • Built multi-agent systems with specialized roles • Applied advanced prompt engineering with few-shot learning and context window optimization • Developed real-time Voice AI agents using Pipecat for low-latency conversational interactions • Implemented streaming speech-to-text → LLM reasoning → text-to-speech pipelines 𝐅𝐮𝐥𝐥-𝐒𝐭𝐚𝐜𝐤 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 • Backend Development: Python, FastAPI for scalable services • Vector Database Integration: Qdrant, Faiss, LanceDB for efficient similarity search • Hybrid Application Development: Cross-platform solutions with Quasar Framework • Mobile Application Development: Expertise with Capacitor and Electron frameworks 𝐂𝐨𝐫𝐞 𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐢𝐞𝐬: - 𝐀𝐈/𝐋𝐋𝐌: Langchain, LangGraph, LlamaIndex, OpenAI, Claude - 𝐃𝐚𝐭𝐚𝐛𝐚𝐬𝐞𝐬: Neo4j, Qdrant, FAISS, LanceDB, Pinecone, Chroma - 𝐕𝐨𝐢𝐜𝐞 𝐀𝐈: Pipecat, Real-time Voice Agents, STT/TTS streaming pipelines - 𝐖𝐞𝐛 𝐂𝐫𝐚𝐰𝐥𝐢𝐧𝐠: FireCrawl, Jina, Puppeteer, Scrapy, Playwright, Beautiful Soup - 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐢𝐧𝐠: Llama Parse, Document Loaders, spaCy, Textacy