Hamdi Barkous

• • Building an AI co-founder SaaS letting non-technical founders ship product changes via chat, voice, and async delegation.

• • Architected a dual-brain agent system: a LiveKit voice agent (GPT-4.5) delegating complex coding tasks to a LangGraph deep agent (Claude) via a shared memory layer.

• • Designed session-durable agent memory in a Daytona sandbox, eliminating conversation summarization and preserving cross-session context.

• • Built a real-time SSE progress pipeline (LangGraph → Zustand) with human-in-the-loop interrupt/resume via the Com mand pattern.

• • Shipped multi-channel surfaces (web, voice, WhatsApp, Telegram) on NestJS 11 + Next.js 16, with Langfuse observability and Stripe billing.

• • Engineered DL&ML models to predict price movements across 500+ financial symbols, achieving a 2+ profit factor in live trading.

• • Built an agentic sentiment analysis framework aggregating 10+ data sources per ticker to generate real-time trading signals.

• • Architected rule-based and LLM-powered trading strategies with built-in explainability for risk management decisions.

• • Created a backtesting engine validated against 5+ years of historical market data, enabling rapid strategy iteration.

• • Streamlined data pipeline architecture, cutting processing time by 60%.