Agentic AI & RAG Data Scientist

Anubhav Srivastava

Bangalore

SUMMARY

+91-999******* LinkedIn GitHub **********.************@*****.*** Data Scientist with 4+ years of experience building AI, Machine Learning, Generative AI, and Agentic AI solutions for banking and enterprise analytics. Expertise in RAG, GraphRAG, MCP, multi-agent systems, NLP-to-SQL, LLM orchestration, and fraud analytics using LangGraph, LangChain, FastAPI, and vector databases. Experienced in developing tool-enabled AI applications, enterprise knowledge retrieval systems, and human-in-the-loop workflows, with a focus on scalability, explainability, and operational efficiency. SKILLS

• Generative AI: LLMs, RAG, Agentic AI, GraphRAG, Prompt Engineering, Fine-tuning, Embeddings, LLM Evaluation (DeepEval), Guardrails, LLM Gateways (Portkey)

• Frameworks & Protocols: LangChain, LangGraph, LlamaIndex, LangMem, MCP, Function Calling, Tool Use, Hugging Face Transformers

• Deployment & LLMOps: FastAPI, REST APIs, LangServe, Docker, Git, CI/CD, Kubeflow, LangSmith (Tracing, Monitoring, Observability)

• Retrieval & Knowledge Graphs: Qdrant, ChromaDB, Pinecone, Neo4j, Hybrid Search, Semantic Search, Reranking, Knowledge Graphs

• Machine Learning: Scikit-learn, XGBoost, Random Forest, Logistic Regression, NLP, PyTorch

• Programming & Cloud: Python, SQL, GCP (BigQuery, Vertex AI, Cloud Storage), AWS (Bedrock, SageMaker) EXPERIENCE

Data Scientist Aug 2024 - Present

HSBC Bangalore

• Designed and deployed production-grade an Agentic AI-powered AML Investigation Copilot using LangGraph, GCP Big Query, FastAPI, and RAG to assist financial crime investigators through explainable, auditable, human-in-the-loop investigation workflows by retrieving AML policies, historical investigation cases, customer profiles, and transaction context for faster investigations.

• Designed production-ready multi-agent workflows with planner, retrieval, reasoning, and summarization agents, leveraging tool calling and semantic search to automate investigation workflows and generate policy-aligned recommendations.

• Built and deployed end-to-end production NLP-to-SQL platform using LLMs and RAG, enabling natural language querying over enterprise banking datasets and reducing dependency on data teams by ~40%, improving query turnaround from hours to seconds.

• Designed and optimized a scalable LLM application using Gemini/OpenAI, FastAPI, BigQuery, Vector DB, and Prompt Orchestration, reducing response latency by ~30% while improving contextual response quality and traced using LangSmith.

• Developed an enterprise RAG pipeline for banking knowledge retrieval by implementing document ingestion, intelligent chunking, embeddings, metadata enrichment, hybrid retrieval, and semantic search over internal documentation.

• Built an LLM-powered information extraction and classification pipeline to convert unstructured fraud complaint data into structured outputs, reducing manual effort by ~70% and improving downstream investigation workflows.

• Periodically retrained machine learning models (XGBoost, Random Forest, Logistic Regression) using Kubeflow for fraud detection and mule account identification, improving detection precision by ~20% while reducing false positives. Data Analyst Dec 2022 - Aug 2024

Capgemini Bangalore

• Built scalable ETL pipelines using Python and BigQuery, improving data processing efficiency by ~30% and delivering analytics- ready datasets for downstream ML/AI applications.

• Developed automated data ingestion workflows across 40+ data sources and built Power BI dashboards for Unsecured Lending, reducing manual effort by ~60% while enabling actionable business insights through automated reporting.

• Designed and maintained campaign analytics pipelines for banking products (Loans, Deposits, FX), aligning SQL-based business logic with stakeholder requirements and reducing reporting turnaround by 20–25%. PROJECTS

MCP-Based Travel Assistant Python, MCP, LangChain, OpenAI Project Link

• Developed an MCP client-server application integrating an OpenAI-powered LLM with a Python-based Travel MCP Server, exposing a search_flights tool for natural-language flight search requests.

• Integrated MCP tools with LangChain's agent framework using STDIO transport, enabling tool discovery, automated tool selection and execution, and result processing through the LLM agent loop. Enterprise Agentic RAG System Google Cloud, LangGraph, FastAPI, Qdrant Project Link

• Built a production-grade Retrieval-Augmented Generation (RAG) system using Python, LangGraph, FastAPI, Groq, Qdrant, and Google Cloud, enabling accurate document-grounded responses over large unstructured datasets.

• Designed and deployed an end-to-end AI pipeline with document ingestion, vector search, reranking, observability, and cloud deployment, improving retrieval quality and scalability for enterprise use cases. Enterprise AI Support Copilot Project Link

• Built an Agentic Enterprise Support Copilot using LangGraph, FastAPI, ChromaDB, and Streamlit for context-aware responses grounded in enterprise documents and customer history.

• Integrated LangMem-based long-term memory, RAG-powered semantic retrieval, and LLM-driven tool calling for customer plan and workload analysis, enabling faster, more consistent, and policy-aligned support workflows for insurance claims. GraphRAG-based Medical Knowledge Retrieval

• Built a GraphRAG pipeline for medical data using NER for entity extraction and Neo4j for knowledge graph construction and storage using Cypher.

• Implemented LLM-based Text-to-Cypher to convert natural-language queries into Cypher queries, retrieve relevant graph context, and generate contextual responses.

Domain-Specific LLM Fine-Tuning for Sales Q&A

• Fine-tuned an open-source LLM using LoRA and QLoRA with Unsloth on a sales-domain Q&A dataset to improve responses to domain-specific queries while reducing fine-tuning resource requirements.

• Prepared and formatted training data and evaluated the fine-tuned model against the base model on representative sales queries to assess response quality.

Swarm AI – Agentic Automation Platform (n8n + LLMs)

• Designed an event-driven, multi-agent architecture in n8n where AI agents collaborate to execute complex workflows (email calendar travel booking), reducing manual effort by ~50%.

• Developed a Multimodal AI assistant with modular workflows, Google Workspace/search integrations, and Telegram + voice interface for end-to-end task automation.

EDUCATION

M.Tech, Nit Rourkela (2022)

B.Tech, JSS Noida (2018)

Contact this candidate