Abdul Rehman

Architected a production-grade Multi-Agentic Financial Advisor serving B2C/B2B users, leveraging Redis for session state and Pinecone for semantic retrieval across financial knowledge bases. Created a document comparison tool with Azure OpenAI and Pinecone, automating cross-document consistency checks and reducing manual validation effort. Deployed a voice-based sales agent with custom STT/TTS pipelines, improving client engagement through natural conversational interfaces. Prototyped an Asset Management Chatbot with curated datasets and preprocessing pipelines, demonstrating feasibility for wealth management use cases. Engineered an on-premise Document Parsing Automation system for banking clients, handling handwritten, English, Urdu, and PDF invoices with field extraction, approver-editable review UI, and ERP (Oracle Apex) integration—powered by Qwen-8B via Ollama. Led technical pre-sales for the GCC region, contributing to multiple AI RFPs and translating client requirements into solution architectures that secured new business. Shipped 15+ AI prototypes across finance, insurance, and enterprise domains, accelerating stakeholder buy-in and use-case validation cycles. Mentored 3 interns on agentic AI architecture, RAG systems, and prompt engineering best practices

Directed end-to-end design and deployment of AI solutions using LangChain agents, custom tool integrations, and prompt engineering. Built semantic similarity classification pipelines processing 1,000+ documents in real time for automated categorization workflows. Generated and indexed vector embeddings using MiniLM-L6-v2 with FAISS, Pinecone, and ChromaDB for scalable knowledge retrieval. Connected LangChain agents with Tavily search APIs, improving information retrieval accuracy and reducing manual research overhead.

Key accomplishments include: Engineered innovative prompt templates and AI agents for natural language processing tasks, resulting in a 25% increase in accuracy and efficiency of AI-driven solutions across multiple projects. Generated and integrated vector embeddings using open-source models like miniLLM with Pinecone, FAISS, and ChromaDB, enhancing data retrieval speed by 30% during complex queries. Developed SQL Agents that streamlined data processing workflows through few-shot learning models, reducing data handling time by 40% while improving decision-making capabilities significantly.

Skills: Natural Language Processing (NLP), Vector Embeddings, SQL Development, Few-Shot Learning, Open-Source Technologies.