Ahmed – LLM, RAG, LangChain

An agentic AI platform for automated credit and compliance investigations, designed to be fully domain-agnostic with modular, swappable use-case packs. The product is currently under active development.

• Built a production LangGraph state-machine orchestrator with nested investigation loops, quality-verify cycles, and targeted section re-runs - ensuring every check completes before a human reviewer decides;
• Designed a domain-blind architecture where all business logic (investigation guides, regulatory rules, report structure) lives in versioned use-case packs, enabling new markets to ship without touching platform code;
• Implemented a multi-stage RAG pipeline using hierarchical chunking, proposition extraction, contextual retrieval (Anthropic technique), hybrid dense+BM25 retrieval with RRF fusion, and BGE cross-encoder reranking- optimized for Arabic financial/regulatory documents;
• Built a citation-first provenance model ensuring every retrieved fact carries a traceable source before it reaches the LLM;
• Designed a guardrail framework (output-shape validators + ReAct LLM review loop) that flags violations without blocking investigation completion;
• Evaluated RAG systems using RAGAS, measuring faithfulness, answer relevancy, context recall, and context precision to ensure retrieval quality and response accuracy.
• Implemented and maintained agent observability using LangSmith, enabling tracking, debugging, and performance analysis of agent workflows.

Co-founded and built an AI-powered platform that helps homebuyers evaluate properties through a conversational AI interface. Led the development of the entire technical stack from scratch through to production launch, focusing on scalable architecture, user experience, and end-to-end system delivery.

• Built a GPT-4.1–powered conversational AI system with custom prompt engineering and tuning to support a domain-specific user experience;
• Designed and implemented a RAG pipeline indexing 400+ property listings using Azure AI Search, including tailored extraction strategies for diverse real estate document types;
• Developed an MCP server for real estate data access, enabling AI assistants to retrieve live property listings, cadastral records, and market analytics;
• Built multi-agent workflows using LangGraph with role-based context engineering, supporting distinct conversation flows for anonymous users and authenticated property owners, with integrated safety guardrails to mitigate hallucinations in legal and financial contexts;
• Integrated external APIs and government data sources (Kadaster BAG, KIK-inzage, RVO, CBS, Altum AI), implementing secure authentication, data orchestration, and optimized caching strategies;
• Delivered a cloud-native Azure architecture, deploying FastAPI services via Azure Container Instances and Azure Functions, leveraging Azure AI Search for vector storage, Azure Front Door for CDN, Azure AD B2C for authentication, and Cosmos DB for persistence;
• Established CI/CD pipelines using Azure DevOps, including automated builds, deployments, and monitoring via Application Insights;
• Contributed across the full-stack, including React frontend development, Python backend engineering, cloud infrastructure setup, DevOps, and product-level design decisions.

The project was an AI-powered, agentic CRM platform designed for startups and SMBs. It focused on automating customer relationship management processes such as lead management, customer communication, and sales workflows using AI agents to improve efficiency and reduce manual effort.

• Architected and built a fully agentic CRM where all user interactions are AI-driven. From lead capture through business plan generation, eliminating traditional form-based UI;
• Designed MCP server + client architecture enabling GPT-4.1 to access customer data, CRM operations, and financial modeling tools through structured tool calling;
• Implemented multi-agent business plan generator using LangGraph: specialized agents for market research, financial projections, competitive analysis, and risk assessment - working in iterative refinement loop until plan meets quality thresholds;
• Built custom financial calculation agent tools for startup-specific metrics, reducing business plan generation time from 2 weeks of manual work to 15 minutes, AI-assisted.

A large-scale digital transformation initiative across financial services and retail domains. My work focused on building end-to-end enterprise architectures for mission-critical platforms, defining integration and data strategies, and ensuring scalable, secure, cloud-native solutions across complex multi-stakeholder environments.

• Owned a core business domain within a next-generation enterprise architecture program, covering agreements and policy management as one of the key functional areas;
• Led enterprise and solution architecture design, including system integration, domain modeling, and data strategy across multiple internal teams and external vendors;
• Supported vendor evaluation and selection processes, ensuring alignment with long-term technical and business architecture goals;
• Managed stakeholder communication and alignment across business and technology teams to ensure consistent delivery of architectural vision;
• Designed PCI-DSS compliant architecture for a digital banking platform, ensuring secure payment and card processing capabilities;
• Defined and led architecture for card management systems and integrations with external financial processing platforms;
• Established architectural direction for back-office systems within a digital banking environment;
• Participated in architecture governance and steering committees for large-scale financial transformation programs;
• Contributed to the design of a large-scale cloud-based retail platform deployed across multiple international markets;
• Owned key domains including appointment management, subscription systems, multi-tenancy, and security architecture;
• Supported the design and integration of both customer-facing (B2C) and internal enterprise (B2E) systems within a unified digital platform;
• Designed solution architecture for personal finance management capabilities within a digital banking ecosystem;
• Led integration design between digital channels and core banking systems, ensuring secure and scalable data flows.