Hassan Elkhatib

AI & Full-Stack Engineer specializing in production Generative AI and agentic LLM systems for the financial / RegTech domain. Build end-to-end pipelines — from RAG and multi-agent orchestration (LangGraph, MCP) to real-time risk surveillance, semantic caching, and developer automation. Strong record of shipping intelligent, data-driven features that cut latency and inference cost while scaling reliably in production.

AI / ML Engineer at Unimade Solutions (2025-09 – Present)

• Risk Surveillance (Gemini, agentic): Architected a four-layer LLM surveillance pipeline — deterministic pre-filter → Gemini Flash screener → Gemini Pro analyst agent (orchestrated via LangGraph + MCP) → decay-weighted scoring store — that monitors live trading activity and automatically flags abusive patterns such as scalping, latency arbitrage, and swap abuse.
• Defined the system's behavioral metric contract, mapping 37 metrics across 9 feature categories to platform APIs, and delivered insights through a dual chatbot + dashboard interface for risk analysts.
• LinkedIn Content Automation: Built an LLM generator–critic pipeline that automates the company's RegTech social presence across Safeguarding, MiFID II, K-Factor, and CRS topics, and ran a structured evaluation that surfaced and corrected critic miscalibration to improve regulatory accuracy.
• GitHub Code-Review Pipeline: Built an automated AI code-review workflow that runs on every pull request, giving the team consistent, standardized review coverage and catching issues before merge.
• Finkit' AI Risk Chatbot: Led development of an AI risk-management chatbot — an intent-routing assistant over live financial risk data — that resolves analyst queries conversationally and removes manual lookup steps.
• Optimized the data layer with a semantic caching layer (serving recurring queries from cached embeddings) over a unified pipeline aggregating three real-time financial sources at 10,000+ records/day, cutting retrieval latency and inference cost.

AI Full-Stack Engineer at Jarvis AI Assistant (2024-12 – Present)

• Orchestrated a multi-model AI assistant that routes across Gemini 2.5 Flash, Gemini 2.5 Pro, and DeepSeek R1, selecting per request to balance latency against answer quality.
• Engineered offline resilience with a locally hosted Gemma 3 fallback, keeping core features available with no internet connection.
• Extended the assistant to multimodal output, adding Flux 1.1 image generation and Veo 3 video generation.

AI Full-Stack Engineer at StudentBuddy (2025-05 – 2025-08)

• Built an academic chatbot using Retrieval-Augmented Generation (LangChain +.NET Core) handling 500+ daily student queries.
• Applied prompt-engineering and retrieval tuning to improve answer accuracy across subjects, and integrated structured + unstructured course data to reduce teachers' manual administrative workload.

M.Sc. in Artificial Intelligence – Lebanese University (2026-06)

B.Sc. in Computer Science – Lebanese International University (2024-06)